A venture fund is not a company that invests money — it is a fixed-life legal contract between people who have money (LPs) and people who pick companies (GPs), wired with a fee-and-profit-share that shapes every decision the GP makes. Understand the plumbing and you can read a fund's incentives, its timeline, and why even good funds look like failures for their first five years.
A venture capital fund is, legally, a limited partnership — in the UK usually a Limited Partnership or a Scottish LP; in the US a Delaware LP. It is a pass-through vehicle, not a trading company: it has no employees, no office, no products. It exists only to hold investments on behalf of its partners and to pass gains and losses straight through to them for tax. This structure is the near-universal default across the industry (NVCA).
Two features define it. First, it is closed-end: investors commit a fixed sum up front, cannot withdraw during the fund's life, and get their money back only as investments are sold. Second, it is a blind pool: LPs commit before knowing which companies the fund will back, betting on the GP's judgement rather than a named portfolio. That is why LP diligence focuses on the team, the thesis, and the track record — there is nothing else to inspect.
A single "firm" (Union Square Ventures, Balderton, Accel) is really three overlapping legal entities: the fund (the LP that holds the assets), the General Partner entity (a separate LP or LLC that legally controls the fund and receives the carry), and the management company (which employs the investors and receives the management fee). Keeping these separate is deliberate — it isolates liability and cleanly routes the two different money streams, fees and carry, to different places.
Limited Partners (LPs) supply the capital. In the UK these are pension funds, insurers, university endowments, sovereign wealth funds, funds-of-funds, and — a UK peculiarity — the government-owned British Business Bank, whose British Patient Capital and Enterprise Capital Funds programmes are among the largest LPs in domestic venture. "Limited" is literal: their liability is capped at what they commit, and in exchange they must stay passive — no involvement in investment decisions, or they risk losing that liability shield.
General Partners (GPs) run everything: sourcing, diligence, pricing, board seats, follow-on decisions, and exits. They carry unlimited liability for the partnership (hence the layer of legal entities to contain it) and are bound by a fiduciary duty to the LPs, codified in the Limited Partnership Agreement (LPA) — the contract that governs fees, carry, the fund's life, permitted investments, and conflict rules. The NVCA model documents are the industry-standard templates that shape these terms.
The single most common beginner error is to assume a £100m fund has £100m in the bank. It does not. LPs sign a capital commitment — a legally binding promise to provide up to that amount when asked. The fund holds almost no cash. As deals close, the GP issues a capital call (a "drawdown" notice), and LPs must wire their pro-rata share, usually within 10 working days.
This "just-in-time" funding exists because idle cash destroys returns. LP performance is measured by IRR (internal rate of return), which is time-sensitive: money that sits uninvested drags the IRR down. So capital is called only as needed — for investments and for fees. Committed capital is the promise; called (or paid-in) capital is what has actually been drawn. Failing to meet a call triggers brutal default provisions in the LPA (forfeiture of some or all of the LP's interest), which is why LPs hold liquidity in reserve for years against calls they cannot precisely predict.
The GP needs to pay salaries, rent, lawyers and travel for years before any investment is sold. The management fee funds this — canonically 2% per year of committed capital. On a £100m fund that is £2m annually to run the firm, called from LPs alongside investment drawdowns.
The critical nuance is the step-down. Charging 2% of committed capital for the full ten years would be indefensible once the fund stops making new investments — you should not pay a manager to actively deploy money they are no longer deploying. So the fee typically holds at ~2% during the investment period, then declines, and often switches from a percentage of committed capital to a percentage of invested cost (the cash still tied up in live companies), which naturally shrinks as companies exit or are written off. Fred Wilson describes Union Square Ventures' schedule as 2.5% for the first two years, stepping down 0.25% a year and reaching zero after year ten (AVC). The point of the taper, in his words, is to push partners to care about profit rather than fee income.
Carried interest — "carry" — is the GP's share of the profits, canonically 20%. This is where GPs are supposed to get rich, and where incentives align: the GP only earns carry if LPs make money (AngelList). Together, "2 and 20" is the shorthand for the whole model.
Carry is paid through a distribution waterfall, the order in which exit proceeds are split:
Note a genuine industry divide: hurdles are standard in private equity and buyout funds but frequently absent in early-stage venture, on the logic that a venture fund targeting 3x+ shouldn't need an 8% floor to prove it earned its carry. Know which regime a given fund uses before reading its terms.
Two further structures matter. Whole-fund ("European") carry pays the GP only after the entire fund's paid-in capital is returned — LP-friendly. Deal-by-deal ("American") carry pays carry on each winning exit as it happens, which is GP-friendly and more common in syndicates and PE than in institutional VC.
Finally, the GP commitment: LPs require the GP to invest its own money into the fund, historically ~1–2% of fund size, so the managers have "skin in the game" and lose alongside LPs if the fund fails. On a £100m fund that is £1–2m of the partners' own capital — a meaningful check on reckless bets.
A venture fund runs on a fixed clock, typically ten years, split into two phases:
The ten-year life almost never holds in practice. Startups take longer to exit than the clock allows, so LPAs grant extensions — usually two one-year options at the GP's discretion, sometimes more with an LP vote. A "ten-year fund" routinely runs twelve to fourteen years. This matters to LPs' liquidity planning: capital committed in 2026 may not fully return until the late 2030s.
Plot a fund's net return over time and it traces the letter J. For the first several years it is negative, then it climbs and (if the fund is good) rises well above the starting line. This is the J-curve, and it is structural, not a sign of failure (Crunchbase).
The early dip has three causes acting together. First, fees come out immediately — the fund is paying 2% a year from day one, so paid-in capital exceeds portfolio value before anything appreciates on paper. Second, losers surface before winners: weak companies fail and get written down within two or three years, while the breakout companies take seven-plus years to mature and are held at cost until a priced round or exit revalues them. Third, illiquid holdings are marked conservatively, so unrealised gains lag reality. The result: measured IRR is typically negative for the first three to five years, then inflects as exits arrive. A savvy LP treats early negative marks as expected physics — and treats a GP who blames the J-curve for genuinely poor performance with suspicion.
Take a £100m fund with a ten-year life and a stepped fee: 2.0% of committed capital during the five-year investment period, then tapering by 0.25% a year through harvest. We compute total fees, how much capital is left to invest, and what that does to returns.
| Year | Phase | Fee rate | Annual fee |
|---|---|---|---|
| 1 | Investment | 2.00% | £2.00m |
| 2 | Investment | 2.00% | £2.00m |
| 3 | Investment | 2.00% | £2.00m |
| 4 | Investment | 2.00% | £2.00m |
| 5 | Investment | 2.00% | £2.00m |
| 6 | Harvest | 1.75% | £1.75m |
| 7 | Harvest | 1.50% | £1.50m |
| 8 | Harvest | 1.25% | £1.25m |
| 9 | Harvest | 1.00% | £1.00m |
| 10 | Harvest | 0.75% | £0.75m |
| Total | £16.25m |
Over the fund's life, LPs pay £16.25m in fees — 16.25% of committed capital, averaging 1.625% a year (already well below the "2%" headline, thanks to the step-down). That leaves only £83.75m to actually put into companies. This is fee drag: roughly one pound in six never reaches a startup.
The consequence for returns is stark. To hand LPs back just their £100m (a 1.0x return of capital), the £83.75m of deployed capital must grow to £100m — a 1.19x gross multiple on invested capital before the fund even breaks even at the LP level. The portfolio has to be worth ~19% more than the cash deployed simply to cover fees.
Now layer in carry. Suppose the £83.75m deployed returns a strong 3.0x — £251.25m of exit proceeds. The waterfall (whole-fund, 20% carry, no hurdle) runs:
So a portfolio that tripled the deployed cash delivers LPs a net multiple of ~2.2x, not 3x. The gap — from 3.0x gross to 2.2x net — is the combined bite of fee drag and carry. This is exactly why LPs distinguish gross from net returns and why the industry rule of thumb holds that a fund must roughly double gross just to give LPs a respectable net result.
Losing 16% of the fund to fees is such a drag that most LPAs permit recycling: the GP may reinvest early exit proceeds and management fees rather than distributing them, so total capital deployed can approach — or, with aggressive recycling, exceed — 100% of committed capital. If our fund recycles ~£12m of early liquidity and fees back into new investments, deployed capital climbs from £83.75m toward ~£96m, and the "£1 in £6 never invested" problem largely disappears. Recycling is usually capped (e.g. up to 120% of commitments) and time-limited to the investment period. When you compare two funds, check the recycling terms — they can matter as much as the headline fee.
Fund-level economics are broadly global, but the UK wraps distinctive tax and institutional machinery around them:
Every number in a fund's structure is an incentive, not just a cost. The 2% keeps the lights on; the step-down stops it becoming a salary; the 20% carry and the GP commitment force the partners to care about exits, not assets; the ten-year clock forces liquidity; and the J-curve guarantees you must judge a fund on a decade, never a quarter.
Most of what your intuition knows about averages is wrong for venture capital. In a normal world — heights, exam scores, daily temperatures — outcomes cluster around a mean and extremes are vanishingly rare. Venture returns do the opposite: they are dominated by a handful of extreme winners, and the "typical" investment loses money. Getting this one fact into your bones changes everything downstream — how many companies you back, how much you own, how you size cheques, and how you keep powder dry. This chapter builds the power law from first principles and then turns it into the concrete arithmetic of portfolio construction.
A normal distribution (the bell curve) arises when an outcome is the sum of many small, independent effects — each nudging the result a little up or down. Additive processes cancel out at the extremes, so you get a tidy hump around the average. But a startup's value is not a sum of small nudges. It is a product of compounding factors: market size × growth rate × retention × margin × time, each multiplying the others over years. Multiplicative processes with any variance produce fat-tailed distributions — most outcomes are small, but the rare large ones are astronomically large. This is the mathematical signature of a power law.
A power law says the probability of an outcome of size x falls off as x raised to a negative exponent (the "alpha"). The lower the alpha, the fatter the tail. Jerry Neumann's analysis in Power Laws in Venture estimates the alpha for VC-backed startups at roughly 1.7–1.97. The practical consequence of an alpha near 2 is startling and worth memorising: within the set of companies that succeed, a single one is likely to return as much as all the others combined. The winner doesn't just beat the pack; it dwarfs the pack.
Why does this happen structurally? Startups attack winner-take-most markets with network effects, economies of scale, and increasing returns to adoption. The company that pulls ahead attracts more capital, talent, and customers, which lets it pull further ahead. There is no force pulling outcomes back toward a mean — the opposite. That reflexive compounding is what manufactures the tail.
The canonical dataset comes from Correlation Ventures, analysed by Seth Levine in Venture Outcomes Are Even More Skewed Than You Think. Across more than 21,000 financings from 2004–2013:
This demolishes the folk heuristic that "a third fail, a third return capital, a third do well." Reality is far more skewed toward zero — and far more dependent on the extreme right tail. Benedict Evans, drawing on Horsley Bridge fund data in Winning and Losing, found the same shape from the fund's-eye view: about half of all investments returned less than the money put in, yet 6% of deals produced at least 10x and those 6% made up roughly 60% of total returns. His most counter-intuitive finding: the best-performing funds did not have fewer failures. They had more — because reaching for the tail means accepting more zeros. Superior returns come from bigger wins, not fewer losses.
Consider a £30m fund that writes a £1m cheque into each of 30 companies (we ignore reserves for now to isolate the effect). Outcomes land like this: 20 companies return 0x (£0), 6 return 1x (£6m total — investors just get their money back), 3 return 3x (£9m total), and 1 returns 50x (£50m). Total proceeds: £0 + £6m + £9m + £50m = £65m, a gross fund multiple of 65 ÷ 30 = 2.17x. Now look at where that came from: the single 50x company produced £50m — 77% of everything the fund returned, and 1.67x the entire fund on its own. Strip out just that one deal and the fund returns £15m on £30m — a 0.5x disaster that loses half its capital, despite 9 of the remaining 29 companies "working". The outlier is not the icing. It is the cake.
That worked example gives us the single most important concept in portfolio design: the fund returner — one investment whose proceeds equal or exceed the entire fund. Because the power law guarantees your result will be dominated by one or two names, every investment must be underwritten to a brutal test: if this works, can it plausibly return the whole fund by itself? If the honest answer is no, the deal cannot move the needle even in its best case, and it is a poor use of a portfolio slot — however "safe" or "reasonable" it looks.
This is why VCs sound obsessed with huge outcomes and seemingly indifferent to solid, modest ones. A company that could become a comfortable £40m business is a fine outcome for its founders and a bad one for a £30m fund's economics if the fund owns only a slice of it. The maths of the tail forces the discipline: you are not trying to be right often; you are trying to be enormously right occasionally.
The fund-returner test connects three levers through one clean identity. The proceeds from a deal are approximately your ownership at exit × the company's exit value. For a deal to return the fund:
ownership at exit × exit value ≥ fund size
Rearranged, this tells you the exit you must believe in for a given ownership stake — and therefore how much ownership you must fight for. The table below applies it to a £30m fund. Note how ownership isn't a vanity metric; it directly sets the difficulty of the bet.
| Ownership at exit | Exit value needed to return the £30m fund | How rare is that? |
|---|---|---|
| 5% | £600m | Very rare — a top-tier outcome |
| 10% | £300m | Rare but achievable |
| 15% | £200m | Plausible target |
| 20% | £150m | Meaningfully more attainable |
At 5% ownership you need to back a £600m company just to get your fund back once; at 20% a £150m outcome does it. Because £150m exits are dramatically more common than £600m exits, higher ownership widens the set of futures in which you win. This is the entire case for concentration.
Ownership, in turn, sets your cheque. Ownership is roughly cheque size ÷ post-money valuation. To own 15% of a company raising at a £10m post-money, you invest £1.5m. So the chain runs: fund size and target return → required exit → target ownership → cheque size → number of cheques the fund can hold. These are not independent choices. Pick any three and the fourth is determined. A £30m fund that wants ~15% ownership with ~£1.5m initial cheques can hold about 20 initial positions before reserves — not 100, and not 5.
Return to our £30m fund but now design it properly. We reserve half the fund (£15m) for follow-ons — a ratio Fred Wilson describes in Reserves, where USV holds back roughly 50% of a fund for future rounds. The remaining £15m of initial capital, at £1.5m per cheque, buys ~15% in 10 companies. Now the follow-on strategy: of those 10, suppose 4 raise a strong Series A. To hold 15% through a dilutive round you exercise your pro-rata right — investing enough to keep your percentage. If you double down hardest on the two that are clearly breaking out, deploying the £15m reserve mostly into them, you protect ownership precisely in the companies most likely to become the fund returner. The reserve isn't a rainy-day buffer; it is how you buy more of your eventual winner before the price explodes.
Reserves are where the power law meets discipline. Wilson's own fund economics post, Allocating Follow-On Capital, makes the point with hard numbers: a portfolio split into thirds (losers, break-evens, 5–10x winners) with equal cheques into every company returns only about 2.2x — not enough after fees and carry. But allocating unequally — small into the losers (say £500k), large into the winners (£3m) — lifts the same portfolio to about 3.7x gross, roughly the level needed to deliver a good net return to investors. Same picks. Different capital allocation. Nearly double the result.
Two subtleties follow. First, follow-on decisions are made with far more information than the initial cheque — you have watched the team execute for one or two years. Yet the temptation is to prop up strugglers ("we've come this far") and under-fund winners (the new price feels expensive). The power law says do the reverse: starve the losers, feed the winners. Second, running out of reserves is a classic unforced error. Wilson notes that early-stage VCs often make too little on their best companies precisely because they lacked the capital to keep buying. A common rule of thumb is to reserve 1x to 2x your initial cheque per surviving company; USV models this with Monte Carlo simulation to keep a ~95% probability of being able to follow on across the whole portfolio.
Reserve planning is impossible without a realistic view of how companies attrit stage by stage. CB Insights tracked 1,100 US seed-funded companies (2008–2010 cohorts) in its Venture Capital Funnel study. The funnel is steep:
| Stage reached | Share of original seed cohort |
|---|---|
| Raised seed (start) | 100% |
| Raised a second round (≈ Series A) | 48% |
| Reached a fourth round (≈ Series C+) | 15% |
| Exited via IPO or M&A | ~30% |
| Became a $1B+ "unicorn" | 1.07% |
Roughly two-thirds of startups stall — they neither exit nor raise the next round. The biggest single cliff is the Series A hurdle, where about half of seed companies fail to graduate. For a seed fund this drives everything: if only ~48% of your bets even reach a priced A, your reserves must be concentrated on that surviving minority, and your initial portfolio must be large enough that the tiny fraction reaching the top of the funnel still contains a fund returner. These are US figures; the shape holds in the UK/EU, though absolute graduation rates are typically lower given a thinner later-stage capital market — the Series A crunch tends to bite harder in Europe.
This is the central design tension, and there is no free lunch. Two coherent philosophies sit at the poles. The spray-and-pray camp (associated with Dave McClure) argues that because you cannot reliably pick the outlier ex ante, you must buy enough lottery tickets — 50, 100, even 500 companies — to be statistically likely to hold one. The high-conviction camp (associated with Peter Thiel) argues the opposite: there simply aren't that many companies you can have deep conviction about, so you should own large stakes in very few, because thin ownership dilutes even a monster exit into irrelevance.
The power law adjudicates via a simple probabilistic frame. If, say, ~2% of startups become fund-returner-scale outcomes, then the chance of holding at least one is 1 − (0.98)^n, where n is the number of independent bets. That rises steeply with portfolio size:
| Number of investments (n) | Chance of holding ≥1 breakout (at 2% hit rate) |
|---|---|
| 10 | 18% |
| 20 | 33% |
| 30 | 45% |
| 50 | 64% |
| 100 | 87% |
More shots means more chances at the tail — that is the diversification argument, and the arithmetic backs it. But each extra name shrinks average ownership (for a fixed fund) and dilutes the payoff when you win, and it dilutes your attention, which for a hands-on lead investor is a real constraint. The resolution most funds reach is not a pure pole but a shape: enough initial bets to catch the tail, then aggressive, unequal reserves to concentrate ownership in the survivors. You diversify to find the winner and concentrate to profit from it. Note the hidden assumption in the table above — independence and a fixed hit rate. A great picker raises the per-deal hit rate (which favours concentration), while a weaker picker relies more on breadth; and in a downturn outcomes correlate, so bets are less independent than the formula assumes.
The mechanics are universal but the constants differ in Europe. Funds are generally smaller than their US peers, so the same ownership targets imply smaller cheques and often smaller "fund-returning" exits — a £150m exit can genuinely return a £20–30m UK seed fund, which is why European funds can rationally back companies that would be immaterial to a large US fund. Two structural features shape angel and micro-fund construction specifically: the UK's SEIS and EIS schemes give income-tax relief (50% and 30% respectively) plus loss relief on qualifying startup investments, which cushions the many zeros and mathematically encourages wider diversification at the earliest stage; and the thinner European growth-stage market means graduation rates past Series A are lower, so reserves and access to later capital matter even more. The BVCA publishes UK-specific performance and activity data worth calibrating against rather than importing US figures wholesale.
Portfolio construction is the power law made operational. Because a handful of extreme winners will dominate your result, you (1) size a portfolio large enough that the tail is likely to appear at all, (2) underwrite every deal to the fund-returner test so each bet could be that winner, (3) fight for enough ownership that being right pays off in exits that actually occur, (4) hold substantial reserves and deploy them unequally to buy more of the winners as they emerge, and (5) plan all of this against honest, stage-by-stage graduation rates. Do this and a portfolio in which most companies fail can still return several times its capital — not despite the losses, but because the structure was built to survive them and harvest the one outcome that matters.
A venture fund reports a handful of numbers, and every one of them is a compression of a decade of decisions into a single figure. Learn to read the five that matter — IRR, MOIC, TVPI, DPI, RVPI — and you can tell in seconds whether a fund is genuinely returning money or merely telling a good story about paper it still holds. Confuse them, and you will mistake optimism for cash.
Every performance metric answers a version of the same question — what did we get back relative to what we put in? — but each frames it differently. The inputs are always the same three streams: paid-in capital (cash LPs have actually wired to the fund, also called contributions or drawdowns), distributions (cash the fund has returned to LPs from exits), and NAV or residual value (the fund's own estimate of what its still-held positions are worth today).
MOIC, TVPI, DPI and RVPI are all multiples — dimensionless ratios like 2.4x. IRR is a rate — a percentage per year. That distinction is the source of most of the mischief that follows. Carta's fund-performance primer and Invest Europe's Investor Reporting Guidelines (the EU/UK reference standard) both treat these five as the canonical set.
The single most useful identity in fund reporting is:
TVPI = DPI + RVPI
Total value is realised value plus unrealised value. That is all this says — but it is diagnostic. A fund with TVPI 3.0x made of DPI 2.4x + RVPI 0.6x has already returned most of its value in cash. A fund with the identical TVPI 3.0x made of DPI 0.2x + RVPI 2.8x has returned almost nothing and is asking you to believe its own valuations. Same headline multiple; wildly different truth. Always decompose TVPI before you react to it.
MOIC and TVPI are close cousins and often used loosely as synonyms, but they differ on the denominator: TVPI divides by paid-in capital (which includes fees and expenses), while deal-level MOIC divides by capital invested in the company. Because roughly 15–20% of a fund's called capital is consumed by management fees and expenses rather than deployed, gross deal MOIC is structurally higher than net fund TVPI. Two numbers, two denominators — never compare one fund's MOIC to another's TVPI.
Every multiple and rate exists in two versions, and the gap between them is where LPs live.
Gross figures measure the performance of the investments — before the GP's management fee (typically 2% of committed capital per year) and before carried interest (typically 20% of profits above a return of capital, sometimes above a hurdle). Net figures measure what the LP actually receives, after both. A fund can report a gross MOIC of 3.5x and a net TVPI of 2.6x; the fee-and-carry wedge ate the difference.
The rule: when a GP quotes a number without saying which, assume it is the flattering one and ask. LPs are judged on net. Benchmarks like Cambridge Associates' pooled indices are reported net of fees, expenses and carry precisely so that LP-relevant comparisons are apples-to-apples.
IRR is the most cited and least trustworthy venture metric. Three structural features make it flatter early-stage funds and make it manipulable.
1. IRR is exquisitely sensitive to timing. Because it is an annualised rate, a given multiple achieved fast produces a spectacular IRR, while the same multiple achieved slowly produces a modest one. A 3x return in 2 years is a ~73% IRR; the identical 3x over 10 years is ~12%. Early-stage funds that get one quick markup or one fast exit can post triple-digit IRRs on tiny amounts of realised value — a number that says almost nothing about the fund's eventual size of outcome.
2. IRR assumes reinvestment at the IRR itself. The maths implicitly assumes every distribution is reinvested at the same high rate — rarely true, which inflates the figure for high-performing early periods.
3. IRR can be engineered with a subscription line. Modern funds increasingly use a subscription credit facility — a bank line secured against LPs' uncalled commitments — to fund deals first and call capital from LPs months later. Because IRR measures return on capital actually outstanding to LPs, delaying the call shortens the time LP money is at work and mechanically lifts IRR, while leaving TVPI, DPI and MOIC completely unchanged (the same cash eventually moves; only the clock changed). This is the cleanest example of an IRR that improves without any real improvement in performance. When you see a strong IRR next to an unremarkable TVPI, suspect timing effects — a subscription line, an early markup, or a quick partial exit — rather than genuine outperformance.
DPI is the metric that cannot be dressed up. NAV is an estimate; the GP marks its own portfolio, and in venture those marks are frequently the price of the last funding round — which may be stale, inflated, or set by a friendly co-investor. RVPI and (through it) TVPI and IRR all inherit that estimation risk. DPI does not. Distributions are wires that either happened or did not. As the saying on LP desks goes, "you can't eat TVPI" — you can only spend DPI.
This is why sophisticated LPs increasingly anchor on DPI, especially for older vintages where the fund has had time to realise. A fund eight years in with DPI 0.3x and RVPI 2.5x is carrying a mountain of unproven paper; the market downturns of 2022–2024 turned many such markups into down-rounds and write-offs, collapsing celebrated TVPIs. DPI is the metric that survives contact with reality.
Plot a fund's net value over its life and you get a J: it dips below the starting line before rising. In the early years, capital is called and fees are charged, but no company has exited and few have marked up — so paid-in exceeds value, TVPI sits below 1.0x, and net IRR is negative. As winners emerge, mark up, and eventually exit, the curve climbs and (in a good fund) finishes well above the line. Fred Wilson's classic AVC note describes this as the standard shape of value creation in venture and private equity.
The J-curve is why you must never judge a fund by its first few years. Bad deals reveal themselves early (write-downs are quick); good deals take years to compound. A vintage-2024 fund showing TVPI 0.8x in 2026 is not failing — it is in the trough. The metrics only become meaningful as the fund matures, which is exactly why benchmarking is done by vintage year.
A £100m fund. It draws capital over four years, returns cash in three tranches, and finishes with some paper still on the books. All figures are net to LPs.
| Year | Capital called (£m) | Distribution (£m) | Cumulative paid-in | Cumulative distributed | NAV at year-end (£m) |
|---|---|---|---|---|---|
| 0 | 25 | — | 25 | 0 | 23 |
| 1 | 25 | — | 50 | 0 | 44 |
| 2 | 25 | — | 75 | 0 | 66 |
| 3 | 25 | — | 100 | 0 | 80 |
| 6 | — | 20 | 100 | 20 | 140 |
| 8 | — | 120 | 100 | 140 | 90 |
| 10 | — | 100 | 100 | 240 | 60 |
Read the trough first. At year 3, all £100m has been called, nothing has come back, and NAV is £80m — below cost, because ~£20m has gone to fees and a couple of early positions were written down. So:
Now read the finish, at year 10:
The IRR intuition. Treat the terminal £60m NAV as if distributed at year 10, giving a net cash-flow stream of −25, −25, −25, −25 (years 0–3), then +20 (yr 6), +120 (yr 8), +160 (yr 10). Solving for the rate that zeroes the NPV of this stream gives a net IRR of roughly 16%. (At a 15% discount rate the NPV is about +£5.4m; at 17% it is about −£5.0m, so the crossing sits near 16%.)
What is "real" here. Of the 3.00x total value, 2.40x is banked cash (DPI) and only 0.60x is still an estimate (RVPI) — a healthy, mostly-realised fund. The ~16% IRR is respectable but far below what the same 3.0x would have shown had the exits come sooner: pull the £120m year-8 distribution forward to year 5 and the IRR jumps past 20%, with TVPI, DPI and MOIC totally unchanged. That single counterfactual is the whole lesson — the multiples measure how much, the IRR measures how fast, and only cash (DPI) is beyond dispute.
An absolute multiple is meaningless without context: a 2.0x from a 2012 vintage (a decade to work, into a huge bull market) is unremarkable; a 2.0x from a 2020 vintage might be exceptional. So performance is benchmarked against the vintage year — the year the fund began investing — and ranked into quartiles against peers of the same vintage. "Top-quartile" is the industry's shorthand for genuine outperformance.
Data providers — Cambridge Associates, PitchBook, Preqin — publish pooled net returns and quartile breakpoints by vintage. As rough, illustrative orders of magnitude for a mature US venture vintage: a top-quartile net TVPI often sits around 2.5x or higher and net IRR around 20%+, while the median lands closer to 1.5–1.8x TVPI and low-double-digit IRR. Breakpoints shift materially by vintage and provider, so always cite the specific benchmark and date rather than a folk figure. The methodology matters too: Cambridge builds its index from GP-reported audited financials, net of fees and carry, capitalisation-weighted by vintage — so read the methodology note before trusting any headline.
| Metric | What it captures | Estimate or fact? | Gameable? |
|---|---|---|---|
| DPI | Realised cash returned | Fact (wires) | Very hard |
| RVPI | Unrealised paper value | Estimate (marks) | Yes — via valuations |
| TVPI / MOIC | Total value per £ in | Part fact, part estimate | Via the RVPI half |
| IRR | Annualised, timing-weighted return | Estimate (uses NAV + timing) | Yes — timing, sub-lines |
Experienced LPs read the five numbers as a system, not a leaderboard:
The through-line: multiples tell you how much, IRR tells you how fast, and the DPI/RVPI split tells you how much of it is real. Master reading the three together and you cannot be sold a paper story dressed up as a cash record.
A startup's life is punctuated by a series of financing events, each one a negotiated exchange of a slice of the company for the cash to reach the next milestone. Understanding this ladder — who writes the cheque at each rung, how big it is, what the company must prove to climb higher, and how the founders' ownership erodes along the way — is the single most useful mental model in venture capital. This chapter builds it from first principles.
Startups are priced by information, and early on there is almost none. A pre-revenue company is a bet on a team and a hunch; a company with £10m of predictable annual recurring revenue is a bet on a spreadsheet. Because risk falls as evidence accumulates, it would be irrational to fund the whole journey at one price. Instead, capital arrives in rounds — discrete injections, each priced higher than the last if things go well. Every round buys the company a runway of typically 18–24 months to hit a set of milestones that de-risk the next, larger cheque.
This staging protects both sides. Investors limit their exposure to any single unproven step and can walk away if milestones are missed; founders avoid selling too much of the company while it is cheap. The names — pre-seed, seed, Series A, B, C — are conventions, not legal categories. "Series A" simply means the first round of preferred stock designated Series A in the company's articles. What actually matters is the stage of evidence the label signals.
A funding round is not "getting money." It is selling a fraction of future ownership at a price set by how much risk you have already removed. The whole game is removing enough risk between rounds that the price per share goes up faster than your ownership goes down.
Pre-money valuation is what the company is judged to be worth before the new money arrives. Post-money is pre-money plus the amount raised. The investors' ownership is simply the amount they put in divided by the post-money. If you raise £2m at an £8m pre-money, the post-money is £10m and the new investors own 2 ÷ 10 = 20%. That 20% is the ownership sold; the existing shareholders keep the other 80%, but their slice of a now-larger pie.
Two other terms recur. An option pool (or ESOP) is equity reserved to hire and reward employees — usually 10–15%, and typically created or "topped up" out of the pre-money at each round, which dilutes existing holders before the new investor even arrives. And in the earliest rounds money often comes not as priced equity but through a SAFE (Simple Agreement for Future Equity) or a convertible note — instruments that postpone setting a price, converting into shares at the next priced round, usually with a valuation cap and/or discount that rewards the early risk-taker.
Bootstrapping, friends & family, and angels. Before any institution is interested, most companies run on founders' savings, revenue, or small cheques from people who know and believe in the founder. Angel investors — wealthy individuals investing their own money — typically write £5k–£50k each, sometimes clubbing together through an angel network or syndicate to assemble £50k–£250k. Valuations are more art than science, commonly £500k–£2m. Founders might sell 10–20%. In the UK this stage is powerfully shaped by tax relief (see below), which is why so many first cheques are structured as SEIS-eligible equity. The milestone that unlocks the next rung: a working prototype and the first signs that someone, somewhere, wants this.
Pre-seed. A relatively new formalisation of the "first institutional-ish" round. Investors are angels, angel syndicates, micro-VCs, and accelerators. The most famous accelerator, Y Combinator, invests a standard $500,000 (a $125k SAFE for a fixed 7% plus a $375k uncapped SAFE) into every company it accepts. Cheques run $100k–$1m; round sizes £250k–£750k in the UK or $0.5m–$1m in the US; pre-money valuations of roughly $3m–$6m. Founders part with 10–15%. The goal of the money is to build a minimum viable product and find early users. Milestone to graduate: a shipped product and initial traction.
Seed. The round where dedicated seed funds and "super angels" lead. Here the company is expected to show the beginnings of product-market fit — real usage, some revenue, a repeatable way of acquiring customers. Per Carta's State of Private Markets, US median pre-money seed valuations reached about $14.9m in 2024, with round sizes commonly $1m–$5m; UK seed rounds skew smaller, often £750k–£2m. Founders typically sell around 20%. Milestone to graduate to Series A: evidence that the growth engine works and can be scaled with more fuel.
Series A. The first "classic" venture round, led by an institutional VC fund that will take a board seat. The company must now show a repeatable, scalable business — often loosely benchmarked at around $1m–$2m of ARR for B2B software, plus strong growth and retention. US Series A round sizes cluster around $8m–$15m at pre-money valuations of roughly $40m–$50m; founders again sell around 20%. The money buys a real go-to-market team. Milestone: prove that spending on sales and marketing produces predictable, profitable growth.
Series B. The "scale it" round. The business model is no longer in question — the question is execution and market size. Growth-focused VCs lead, round sizes commonly run $20m–$40m, and per Carta, median Series B pre-money valuations sat above $100m through 2024. The company is typically at $5m–$10m+ ARR. Ownership sold drops to 15–20% because the company is now valuable enough that a big cheque buys a smaller slice. Milestone: market leadership in sight and a credible path to much larger scale.
Series C and growth rounds. Now the company raises to dominate its market, expand internationally, acquire competitors, or build new product lines. Investors broaden to growth-equity funds, crossover funds (which invest both privately and in public equities), and late-stage arms of large VCs. Carta put the median Series C pre-money valuation at $354.5m in Q3 2024. Round sizes routinely exceed $50m; ownership sold falls to 10–15%. Milestone: proven, efficient scale and a believable route to profitability.
Late-stage / pre-IPO. The final private rounds (Series D, E and beyond) are raised by companies worth $1bn+ — the "unicorns." Investors include crossover funds, sovereign wealth funds, and private-equity firms. Cheques and rounds run into the hundreds of millions; ownership sold can be under 10%. The purpose is to reach the metrics — predictable growth, defensible margins, robust governance — that public markets or a strategic acquirer will pay for. Milestone: IPO-readiness.
Exit. The endgame that makes the whole chain rational. A trade sale (acquisition by a larger company) or an IPO (listing shares on a public market) turns illiquid equity into cash or freely tradable stock. This is the moment every prior investor was underwriting: venture returns are driven by a small number of large exits, not by the many companies that stall or fail along the way.
| Stage | Typical raise | Typical pre-money | Lead investor type | Key milestone to reach it |
|---|---|---|---|---|
| Angel / F&F | £50k–£250k | £0.5m–£2m | Angels, syndicates | Prototype + a believable founder |
| Pre-seed | £250k–£750k / $0.5m–$1m | $3m–$6m | Micro-VCs, accelerators | MVP shipped, first users |
| Seed | $1m–$5m (UK £0.75m–£2m) | ~$15m (US median) | Seed funds, super angels | Early product-market fit |
| Series A | $8m–$15m | $40m–$50m | Institutional VC | Repeatable, scalable revenue |
| Series B | $20m–$40m | $100m+ | Growth VC | Scaling GTM, market share |
| Series C+ | $50m+ | $350m+ | Growth equity, crossover | Market leadership, path to profit |
| Late / pre-IPO | $100m+ | $1bn+ | Crossover, PE, sovereign | IPO-ready metrics |
Treat these as orders of magnitude, not laws. They move with the macro cycle — 2021 valuations were far higher than 2023's trough — and vary by sector (deep tech and biotech burn more and raise larger, earlier).
Dilution is the mechanism that quietly transfers ownership from founders to investors over a company's life. Each round multiplies every existing holder's stake by pre-money ÷ post-money. Follow a fictional UK SaaS company, Meridian, whose two founders start owning 100%.
The lesson is the whole point of venture finance in one arc: the founders' percentage fell from 100% to 48%, but the value of their stake rose from nothing to roughly £48m, because each round grew the pie far faster than it shrank their slice. Dilution is not something to avoid — it is the price of building something big. (This simplified trace ignores option-pool top-ups, which would typically shave a few more points off founders and early investors at each round.)
Owning a smaller share of a much larger, more valuable company is the desired outcome. Founders who cling to percentage and refuse to dilute usually end up with 100% of something small. The arithmetic to watch is not "how much am I giving up" but "does the new money raise the whole valuation by more than the ownership it costs me."
Most companies do not make it up the whole ladder — and the statistics matter because they define how VCs price risk. The hardest single step is seed to Series A. Carta's cohort data shows the two-year conversion rate collapsing from around 30% for the 2018 seed cohort to roughly 15% for the 2022 cohort — the so-called "Series A crunch." Even over four years, only around half of a strong seed cohort reaches an A. Once past that bottleneck, progression improves markedly: historically around 60% of Series A companies reach Series B, and similar rates apply from B to C and C to D. The implication for founders is stark — raise enough at seed to genuinely clear the Series A bar, because the market will not extend infinite patience.
The ladder is universal but the funding environment is not. Three features distinguish the UK and EU.
Tax-advantaged early rounds. The UK runs two of the world's most generous startup-investment incentives, and they profoundly shape angel and seed rounds. Under the Seed Enterprise Investment Scheme (SEIS), a company can raise up to £250,000, and investors receive 50% income-tax relief on up to £200,000 invested per year, plus capital-gains exemption if shares are held three years. The Enterprise Investment Scheme (EIS) extends this to larger raises — up to £10m a year (more for knowledge-intensive companies) — with 30% income-tax relief on up to £1m per investor (£2m if the extra goes to knowledge-intensive firms). The relief rates are summarised on GOV.UK's tax relief for investors page. Because a 50% rebate roughly halves an angel's downside, SEIS/EIS eligibility is often the first thing a UK angel asks about — it effectively subsidises the riskiest rungs of the ladder and is a big reason the UK has a deep angel base.
Angel networks and a state-backed anchor. UK early-stage capital flows heavily through organised angel networks and syndicates, alongside the British Business Bank, the government-owned economic-development bank that co-invests through funds and programmes to crowd in private capital. Its Small Business Equity Tracker and Small Business Finance Markets reports are the canonical data on UK equity finance.
The growth-stage gap. The UK and Europe are strong at seed and Series A but visibly thinner at the later, larger rounds. The British Business Bank's own analysis notes that the funding gap with the US has widened, and that intermediaries perceive shortages in growth-stage equity in particular. The practical consequence: European companies that reach Series C and beyond frequently raise from US crossover funds, or relocate their centre of gravity to access American growth capital and public markets. Closing this gap — via vehicles like British Patient Capital and pension-fund reforms — is an active policy priority. For a founder, the takeaway is that the ladder gets structurally harder to climb the higher you go if you stay purely domestic.
When someone puts money into a startup, the deal has to answer one deceptively hard question: what fraction of the company does that money buy, and when is that fraction fixed? Every instrument in this chapter is a different answer to that question. Some fix the price today (priced equity). Some deliberately defer the price to a later, better-informed moment (SAFEs, convertible notes, and the UK's Advance Subscription Agreement). Understanding the trade-offs — speed versus certainty, and, in the UK, tax relief versus flexibility — is the core of financing fluency.
A priced round is the classical way to raise: the company and investors agree on a valuation, and shares are issued at a fixed price per share on the spot. The price comes from a pre-money valuation — what the business is agreed to be worth before the new cash arrives. Add the investment and you get the post-money valuation. The investor's ownership is simply their cheque divided by the post-money.
Example: a company raises £2m at a £8m pre-money valuation. Post-money is £10m, so the new investors own £2m ÷ £10m = 20%. If there were 5,000,000 shares before the round, the price per share is £8m ÷ 5,000,000 = £1.60, and the company issues 1,250,000 new shares (£2m ÷ £1.60) to the investors.
Crucially, priced rounds almost never issue plain ordinary shares to venture investors. They issue preferred shares (in the UK often called "preference shares" or, colloquially after US practice, "preferred"). Preferred shares carry a bundle of contractual rights that ordinary shares don't:
Priced rounds are thorough and expensive. Lawyers negotiate a shareholders' agreement and articles of association; legal fees of £15k–£50k+ per side are normal. That is worthwhile at Series A and beyond, where cheques are large. It is disproportionate for a founder trying to close £50k from an angel next week. That mismatch is exactly what deferred-price instruments were invented to solve.
The convertible note — in the UK a Convertible Loan Note (CLN) — is the oldest deferral instrument. Legally it is debt: the investor lends the company money, and the loan is designed to convert into shares at the next priced round rather than be repaid in cash. Because it is a loan, it has debt-like features:
The maturity date and the repayment obligation are the CLN's weaknesses: they hang a potential liability over the company. They are the reason the SAFE was created.
The SAFE — Simple Agreement for Future Equity — was introduced by Y Combinator in 2013 to strip the debt baggage out of the convertible note. A SAFE is not a loan: there is no interest, no maturity date, and no repayment obligation. It is a contract giving the investor the right to shares in the future, when a priced round happens. The investor pays now; the company owes shares later. If the round never comes, the SAFE simply sits there (it converts on other trigger events like an acquisition). This makes it faster, cheaper, and founder-friendlier than a note.
SAFEs and CLNs use two economic levers to reward early investors for taking on early risk and to set their eventual price. Both are described well in YC's own explainer on SAFEs and priced rounds:
Most SAFEs have both a cap and a discount, and on conversion the investor gets to use whichever produces the lower price per share (and therefore more shares) — never both stacked together. A third lever, MFN ("most favoured nation"), has no cap or discount of its own; instead it promises the holder that if the company later issues a SAFE on better terms, this earlier investor can upgrade to those terms. MFN-only SAFEs are used when neither side wants to argue about a valuation yet.
A pro-rata side letter is a separate document, not part of the SAFE itself, granting the investor the right to invest again in the priced round to maintain their ownership percentage — the deferred-instrument equivalent of the pro-rata right that priced-round preferred shares carry.
The original 2013 SAFE was a pre-money SAFE: its cap referred to the company's value before the new priced-round money. The problem was dilution accounting. If a founder sold several pre-money SAFEs, each holder's eventual percentage depended on how many other SAFEs were sold and how they interacted — nobody could state their ownership until the whole stack converted at once. Founders routinely sold far more of the company than they realised.
In 2018 YC replaced it with the post-money SAFE, where the cap is measured after all SAFE money is counted (but still before the new priced-round investment dilutes everyone). The payoff is arithmetic clarity: a £500k post-money SAFE at a £10m post-money cap is exactly 5% of the company, full stop, knowable the day it is signed. The trade-off is that this certainty for the investor comes at the founder's expense — post-money SAFEs push more of the future dilution onto the founder, because SAFE holders no longer dilute each other. The lesson for founders: track the running total of SAFEs sold, because each one is a fixed, non-diluting claim on the cap table.
Here the UK diverges sharply from the US, and the reason is tax. Britain runs two of the world's most generous startup investment incentives: the Seed Enterprise Investment Scheme (SEIS) and the Enterprise Investment Scheme (EIS). SEIS gives an individual investor 50% income tax relief on up to £200,000 invested per year; EIS gives 30% relief on up to £1m (or £2m into knowledge-intensive companies). Both add capital gains exemptions and loss relief. For a UK angel, these reliefs can be worth more than the upside on the shares themselves — so any instrument that breaks SEIS/EIS eligibility is close to unusable.
This is the fatal problem with SAFEs and convertible notes in the UK. SEIS/EIS relief requires the investor to subscribe for new ordinary shares, with money fully at risk, and — critically — the shares must be issued in a way HMRC recognises. A convertible note is debt, which disqualifies it outright: money lent is not money subscribed for shares. A standard SAFE is a contract for future shares whose terms (potential refund, indefinite deferral) generally fail HMRC's conditions too. So the American toolkit largely does not qualify for UK tax relief.
The Advance Subscription Agreement (ASA) is the UK's purpose-built answer: a deferred-price instrument engineered to preserve SEIS/EIS eligibility. SeedLegals, which productised it as "SeedFAST", explains the mechanics in its ASA guide. To qualify, an ASA must obey rules that flow directly from HMRC's requirements:
Within those constraints, an ASA can still carry a valuation cap and/or a discount, working exactly as they do in a SAFE. Discounts are typically 10–20%; SeedLegals' data across thousands of ASAs shows most use either no discount or a 20% discount, and about half include a cap. Caps and discounts are often reserved for foreign investors who cannot use SEIS/EIS anyway, and so want a different form of reward — see SeedLegals' note on choosing deal terms.
The moment of truth for any deferred instrument is the next priced round. The instrument converts into shares, and the arithmetic decides how many. The investor's price per share is set by two candidate prices — one from the cap, one from the discount — and the investor uses whichever is lower (giving them more shares). Let's work it fully.
An angel invests £500,000 via an ASA (a UK convertible works identically here) with a £5m valuation cap and a 20% discount. Eighteen weeks later the company closes a Series A. The round is priced at a £10m pre-money valuation, and the company has 5,000,000 shares outstanding immediately before the round. So the Series A price per share that new investors pay is £10,000,000 ÷ 5,000,000 = £2.00 per share.
Path 1 — the discount. The ASA holder pays 20% less than the round price: £2.00 × (1 − 0.20) = £1.60 per share. Their £500,000 buys 500,000 ÷ 1.60 = 312,500 shares.
Path 2 — the cap. The cap says: convert as if the company were worth no more than £5m. The cap price per share is the cap divided by the pre-round share count: £5,000,000 ÷ 5,000,000 = £1.00 per share. Their £500,000 buys 500,000 ÷ 1.00 = 500,000 shares.
Which wins? The investor takes the lower price: the cap's £1.00 beats the discount's £1.60. They convert at £1.00 and receive 500,000 shares — 60% more shares than the discount would have given. The cap wins because the round (£10m pre-money) was priced at double the cap (£5m); whenever the round valuation runs well above the cap, the cap dominates. Had the Series A instead priced at, say, a £5.5m pre-money (£1.10/share), the discount price (£0.88) would beat the cap price (£1.00), and the discount would win.
The intuition is worth internalising: the discount rewards you for being early by a fixed margin; the cap rewards you disproportionately when the company's value jumps. The cap is your friend in a breakout success; the discount is your floor of reward in a modest one. Because the investor always gets the better of the two, having both is strictly better for them — and correspondingly more dilutive for the founder.
| Instrument | Legal nature | Typical stage / user | UK SEIS/EIS? |
|---|---|---|---|
| Priced equity (preferred) | Shares issued now | Series A onward; lead VCs writing large cheques | Preferred shares usually break SEIS/EIS (special rights); founders use ordinary shares for SEIS/EIS rounds |
| Convertible note / CLN | Debt, converts to equity | Bridges between rounds; US-style seed | No — it is debt |
| SAFE (post-money) | Contract for future equity | US pre-seed/seed; YC companies; global accelerators | Generally no |
| ASA (SeedFAST) | Irreversible subscription, converts ≤6 months | UK pre-seed/seed; angels wanting tax relief | Yes, if the ≤6-month longstop and no-refund rules are met |
The practical decision tree for a UK founder is short. Raising quickly from UK angels who care about tax relief? Use an ASA. Raising from foreign or institutional investors who don't need SEIS/EIS, or bridging to a round more than six months away? A convertible note or SAFE may fit better, at the cost of tax relief. Raising a large, structured round with a lead investor? Do a full priced round. And whatever you sign, keep a running tally of every cap and discount you have promised — deferred instruments are invisible on the cap table until they convert, and that is precisely when founders discover they gave away more than they thought.
A cap table is the ledger of who owns your company — and dilution is the arithmetic of what happens to that ownership every time you sell a slice to raise money. Founders who don't read the maths sign away far more than they think, usually on a single innocuous-looking line: the option pool. This chapter builds the cap table from first principles and then runs a full three-round simulation so you can watch a founding team's stake fall from 100% to 36% — and see exactly where each point went.
A capitalisation table (cap table) is simply a list of every security a company has issued and who holds it. In its earliest form it is a spreadsheet with one row per shareholder and columns for share count and percentage. As the company raises money it grows rows (new investors, converted notes, granted options) and columns (one per financing round). Every equity negotiation, exit payout, and employee offer is ultimately settled by reading this document, so precision matters more here than almost anywhere else in a startup.
The cap table answers three questions: who owns what, what they paid, and what they get in an exit. Get the first wrong and you mis-price a round; get the third wrong (liquidation preferences, covered in the term-sheet chapter) and founders can walk away with nothing from a headline-good sale.
The single most common beginner error is treating a share count as if it were a fixed ownership stake. It isn't. Ownership percentage is your shares divided by the total shares that exist — and the denominator keeps growing. Your share count almost never falls, yet your percentage falls at every round because new shares are created and handed to new owners.
| Holder | Shares | Total shares | Ownership % |
|---|---|---|---|
| You (at founding) | 4,000,000 | 7,000,000 | 57.1% |
| You (after Seed) | 4,000,000 | 10,000,000 | 40.0% |
| You (after Series A) | 4,000,000 | 14,550,000 | 27.5% |
Your 4,000,000 shares never moved. Your stake fell by nearly half. That gap between "shares" and "percentage" is the whole subject of this chapter.
Three denominators exist and they are not interchangeable. Quoting a percentage against the wrong one is how founders get quietly short-changed.
| Category (post-Series A snapshot) | Shares |
|---|---|
| Authorized (constitutional ceiling) | 20,000,000 |
| Issued & outstanding (founders + investors + ASA) | 12,550,000 |
| Options granted (outstanding, unexercised) | 1,000,000 |
| Option pool remaining (available to grant) | 1,000,000 |
| Fully-diluted total | 14,550,000 |
| Unissued / unallocated authorized | 5,450,000 |
Always compute your ownership on a fully-diluted basis. A founder told they own "50%" of issued shares may own far less fully-diluted once a fat option pool and two SAFEs are counted — and it is the FD number that governs an exit.
A priced round has two valuations. Pre-money is what the company is agreed to be worth before the new cash arrives. Post-money is pre-money plus the new investment. They differ by exactly the cheque size, and the relationship that ties them to the cap table is the price per share:
That last identity is worth memorising: an investor's percentage is simply their cheque divided by the post-money valuation. Put £1m in at a £5m post-money and you own 20%, regardless of the share count. The Holloway Guide to Raising Venture Capital frames the whole exercise plainly: raising money means selling part of the company. Pre-money is the price tag on the part you keep.
Seed round: £1,000,000 invested at a £4,000,000 pre-money. Post-money = £5,000,000. The founders hold 7,000,000 shares and the round also creates a 1,000,000-share option pool (see below), so pre-money FD = 8,000,000 shares. Price per share = £4,000,000 ÷ 8,000,000 = £0.50. The investor's shares = £1,000,000 ÷ £0.50 = 2,000,000. Post-money FD = 10,000,000 shares. Investor ownership = 2,000,000 ÷ 10,000,000 = 20% = £1m ÷ £5m. Every figure reconciles.
Startups reserve shares to grant to future employees — the option pool (or Employee Share Option Pool, ESOP, in UK parlance). These are real fully-diluted shares even before anyone is hired, because they will be issued. Fred Wilson's rule of thumb is to size the pool to fund actual hiring plans to the next round, which lands around 7.5–10% at seed rather than a horse-traded number.
Here is the sneakiest term in venture finance. Investors almost always insist the pool be created — or topped up — out of the pre-money, and they count it in the pre-money share base when pricing. The effect: the pool dilutes the founders alone, while leaving the investor's percentage untouched. Wilson calls it exactly what it is — "just another way to lower the price". In his example a £4m pre-money with a 15% pool carved out is economically identical to a £3.25m pre-money with none.
Same seed cheque (£1m at £4m pre), founders holding 7,000,000 shares. Compare no pool against a 10% pool carved from the pre-money.
| Holder | No pool | 10% pool, carved pre-money |
|---|---|---|
| Founders | 80.0% | 70.0% |
| Seed investor | 20.0% | 20.0% |
| Option pool | 0.0% | 10.0% |
The investor gets 20% either way — their cheque still divides by the same post-money. The entire 10-point pool comes out of the founders' 80%. Had the pool instead been created from the post-money (shared pro-rata by everyone), founders would sit at ~72% and the investor at ~18%. The negotiating lever is therefore not "how big is the pool" but "whose side of the pre/post line does it sit on." Push to size the pool honestly to your real hiring plan, and to state pre-money net of the pool so nothing is hidden.
Before a priced round, founders often raise on a SAFE (Simple Agreement for Future Equity — the Y Combinator standard, US) or, in the UK, an ASA (Advance Subscription Agreement — SeedLegals' SeedFAST). Both take cash now for shares issued later, deferring the valuation fight to the next round. They are invisible on the cap table as shares until they convert — which is why they must always be tracked on a fully-diluted basis in the meantime.
Two terms set the conversion price. A discount (typically 10–20%) lets the holder convert below the round price, rewarding early risk. A valuation cap sets a maximum valuation at which their money converts, so if the round prices above the cap they get a lower price and more shares. The holder converts at whichever gives the better (lower) price. UK note: for an ASA to preserve SEIS/EIS relief, HMRC requires conversion within six months and no discount/redemption features that break the rules — a real constraint on how founders structure the bridge.
A £500,000 ASA with a £10,000,000 cap and a 20% discount converts at Series A, where the round price is £2.00/share and the pre-money FD base is 11,000,000 shares. Cap price = £10,000,000 ÷ 11,000,000 = £0.909. Discount price = £2.00 × 0.80 = £1.60. The cap wins. ASA shares = £500,000 ÷ £0.909 = 550,000. The ASA holder gets shares as if they'd bought at a £10m valuation, not the £22m headline — the reward for going in early and blind.
Now we run it end to end. Two founders incorporate with 7,000,000 shares. They raise a Seed round, then a £500,000 ASA that converts at Series A, then Series A and Series B — each with an option-pool top-up carved from pre-money, and with the Seed investor exercising pro-rata rights at Series B to defend their stake. Share counts first:
| Holder | Post-Seed | Post-Series A | Post-Series B |
|---|---|---|---|
| Founders | 7,000,000 | 7,000,000 | 7,000,000 |
| Seed investor | 2,000,000 | 2,000,000 | 2,500,000 |
| ASA / SAFE holder | — | 550,000 | 550,000 |
| Series A investor | — | 3,000,000 | 3,000,000 |
| Series B investor | — | — | 3,000,000 |
| Option pool (granted + available) | 1,000,000 | 2,000,000 | 3,500,000 |
| Fully-diluted total | 10,000,000 | 14,550,000 | 19,550,000 |
The round mechanics: Seed — £1m at £4m pre, price £0.50, 10% pool carved pre-money. Series A — new lead invests £6m at £2.00/share (3,000,000 shares), the pool is topped up by 1,000,000 pre-money, and the ASA converts into 550,000 shares. Series B — new lead invests £18m at £6.00/share (3,000,000 shares), the pool is topped up by 1,500,000 pre-money, and the Seed investor exercises pro-rata, buying 500,000 shares for £3m to defend their percentage. Now the same table as ownership percentages — the number that actually matters:
| Holder | Incorp. | Post-Seed | Post-Series A | Post-Series B |
|---|---|---|---|---|
| Founders | 100.0% | 70.0% | 48.1% | 35.8% |
| Seed investor | — | 20.0% | 13.7% | 12.8% |
| ASA / SAFE holder | — | — | 3.8% | 2.8% |
| Series A investor | — | — | 20.6% | 15.4% |
| Series B investor | — | — | — | 15.4% |
| Option pool | — | 10.0% | 13.7% | 17.9% |
| Price per share | — | £0.50 | £2.00 | £6.00 |
| Post-money valuation | — | £5.0m | £29.1m | £117.3m |
Read the founder row top to bottom: 100% → 70% → 48% → 36%. That arc is normal and, as Fred Wilson argues in Employee Equity: Dilution, healthy — because the value of that shrinking slice is exploding. The founders' 35.8% at Series B is worth roughly 0.358 × £117.3m ≈ £42m, versus their 70% of a £5m company (£3.5m) at Seed. A smaller share of a much larger pie. Dilution only hurts when valuation doesn't grow to compensate — a "down round," where you sell more shares at a lower price and the percentage loss buys you nothing.
Two term-sheet clauses (covered fully in the term-sheet chapter) show up directly as cap-table arithmetic. Pro-rata rights let an existing investor buy enough of a new round to hold their percentage steady. In the simulation, the Seed investor without pro-rata would have drifted to ~10.5% by Series B; by writing a £3m cheque they held 12.8%. Pro-rata is not free — it costs fresh capital — but it lets conviction investors defend their winners, and it comes out of the allocation new leads wanted, so it is often contested.
Anti-dilution protection defends the price an earlier investor paid if a later round prices lower. In a down round, a "full-ratchet" clause re-prices all of the earlier investor's shares to the new, lower price — issuing them extra shares for free and diluting founders hard. The gentler and far more common weighted-average variant adjusts only partially, in proportion to how many shares the down round issued. Either way the mechanism is the same: bonus shares appear on the cap table for the protected investor, and the founders' row absorbs them. It is the reason a down round is doubly painful for founders — lower price and a ratchet firing at once.
A term sheet is the two-page document where a venture financing is actually decided. The long-form legal agreements that follow — the shareholders' agreement, the articles, the subscription agreement — mostly just render its bullet points into enforceable prose. Master the term sheet and you can read any deal; miss a clause here and you will feel it years later at exit, when the money is finally divided and it is far too late to renegotiate.
A term sheet is a short summary of the proposed terms on which an investor will put money into a company. Its defining legal feature is that it is mostly non-binding. Signing it does not commit either side to complete the deal. It is a statement of intent that frames the negotiation and instructs the lawyers who draft the real documents.
"Mostly" is the load-bearing word. A handful of clauses are binding even though the rest is not, and they are always spelled out as such: the no-shop / exclusivity clause, the confidentiality clause, and usually governing law and costs. Everything else — valuation, preferences, board seats — is a good-faith outline. This asymmetry matters: the binding parts overwhelmingly protect the investor, while the economics that protect the founder remain provisional until closing.
Every term on the sheet resolves into one of two dimensions. Economics govern who gets how much money, and when — valuation, the option pool, liquidation preferences, anti-dilution, pro-rata rights. Control governs who gets to decide things — board composition, protective provisions (veto rights), drag-along, and founder vesting. A deal can be generous on economics and brutal on control, or vice versa. Reading a term sheet well means sorting each line into the right bucket and asking, for each, who does this favour, and when does it bite?
The market has largely standardised these documents. In the US the canonical reference is the NVCA model legal documents; in the UK it is the BVCA / UK Private Capital model documents, whose "Summary of Terms" is the British term-sheet equivalent. At the earliest stage, many rounds skip the priced term sheet entirely and use a Y Combinator SAFE or a convertible note, which defer these questions to the next priced round.
The liquidation preference determines the order and amount in which people are paid when the company has a "liquidity event" — a sale, merger, or wind-down. Note the VC-specific meaning: as Brad Feld emphasises in his term-sheet series, "liquidation" here means any exit, not just bankruptcy. It is the single term that most reshapes who gets what, and it is where founders lose the most value without noticing.
It has three moving parts:
Non-participating preferred forces a choice: at exit the investor takes either the preference or converts to common and takes their ownership percentage — whichever is greater, but not both. Participating preferred ("double dip") takes the preference and then also the ownership percentage of what is left. 1x non-participating is the founder-friendly market standard in strong rounds; participating preferred tilts every outcome toward the investor and bites hardest at mediocre exits.
Take a clean cap table. An investor puts in £2m for 20% of the company, at a £10m post-money valuation. Founders and employees hold the other 80%. Now the company sells — not for a fortune, but for £8m. This is the mediocre exit that reveals what the preference actually does. (Assume a simple 1x preference and ignore the option pool for clarity.)
1x non-participating. The investor compares two paths. Take the preference: £2m. Or convert to common and take 20% of £8m: £1.6m. £2m is greater, so they take the £2m preference and stop. Founders receive the remaining £6m. The investor would only convert to common once 20% of the exit exceeds £2m — i.e. above a £10m sale. Below that, the preference is worth more; above it, converting is.
1x participating. The investor takes their £2m preference off the top first. That leaves £6m, which is shared pro-rata. The investor's 20% of that £6m is a further £1.2m. Their total is £3.2m, and founders receive £4.8m. Same company, same price, same 20% stake — but participation has moved £1.2m from the founders' pockets to the investor's.
| Term (£2m in, 20% stake) | Exit at £8m — investor / founders | Exit at £30m — investor / founders |
|---|---|---|
| 1x non-participating | £2.0m / £6.0m | £6.0m / £24.0m |
| 1x participating (uncapped) | £3.2m / £4.8m | £7.6m / £22.4m |
| 1x participating, 2x cap | £3.2m / £4.8m | £6.0m / £24.0m |
Read the £30m column to see two further mechanics. First, non-participating preferred simply converts to common once the company does well — 20% of £30m is £6m, far more than the £2m preference, so the investor abandons the preference. Second, the cap forces a conversion: a 2x-capped investor is limited to £4m of participation, but converting to common yields £6m, so they convert and the cap effectively neutralises participation on the upside. This is exactly why founders who cannot avoid participation should fight hard for a low cap — it turns participating preferred back into ordinary preferred whenever the outcome is good.
Preferences stack across rounds. By Series C there may be three layers of preference sitting ahead of common. In a modest exit — the most common outcome in venture — the preference stack can consume the entire sale price, leaving founders and employees with nothing while investors are made whole. Always model your cap table against a disappointing exit, not just the dream one.
Anti-dilution protects an investor if the company later raises money at a lower price per share than they paid — a "down round". It adjusts the rate at which their preferred shares convert into common, effectively handing them extra shares for free to compensate for the drop. It only bites in a down round; in flat or up rounds it never triggers. Feld's anti-dilution post lays out the two flavours.
The difference is large. If a company with millions of shares issues a small down-round tranche, full ratchet can re-price the entire earlier round as though it had all been sold at the low price; weighted average barely moves the needle. Founders should treat "broad-based weighted average" as the default and "full ratchet" as a red flag worth escalating.
Pro-rata rights (called pre-emption rights in UK usage) give an existing investor the right — not the obligation — to invest in future rounds to maintain their ownership percentage. If they own 20% and you raise a new round, they may buy 20% of the new shares before outsiders are offered any. This favours the investor: it lets winners double down on their best companies and is one of the main ways venture funds generate returns. It matters to founders when a fund insists on a super pro-rata right (more than their percentage), which can crowd out new lead investors in the next round and complicate fundraising. In the UK, statutory pre-emption rights exist by default under the Companies Act and are usually modified by the articles and shareholders' agreement.
The board of directors is where real operational control lives — it hires and fires the CEO, approves budgets, and signs off major decisions. Control of the company is not the same as owning most of the shares; a founder can own 70% and still lose control of the board. Term sheets specify board size and who appoints each seat. A common early-stage structure is a 3-person board: one founder/common director, one investor director, and one mutually-agreed independent director who holds the swing vote. A 2-2-1 structure appears at Series A.
The tell is who controls the majority of seats. As long as founders (plus a friendly independent) hold the majority, they retain board control. Once investors control the board — often before they own a majority of the equity — they can, in the extreme, replace the founder-CEO. Founders should track board composition across rounds as carefully as they track dilution.
Protective provisions are a list of actions the company cannot take without the consent of the preferred shareholders (or their director), regardless of the board vote. They are veto rights, not affirmative powers — investors cannot force actions, but they can block them. Typical items: selling the company, raising a new round, issuing senior stock, changing the articles, taking on significant debt, or changing the size of the board. In UK term sheets these appear as investor consent matters and, per SeedLegals, are among the most hotly negotiated terms.
These favour the investor and are reasonable in moderation — a minority investor legitimately wants to block a founder from, say, selling the company cheaply to a friend. They become a problem when the list is long enough to paralyse ordinary operating decisions, or when a small investor gains a veto over a sale that the majority wants. Founders should push to keep the list short and to set consent thresholds by class (a majority of preferred) rather than granting any single investor a personal veto.
These two clauses govern what happens when the company is sold and not everyone agrees.
The option pool (or ESOP) is equity reserved for future employees. Its size is an economic term, and its placement in the valuation is one of the most common ways founders are quietly diluted — the "pre-money shuffle".
The trick is this: investors insist the new option pool be created out of the pre-money valuation, before their money goes in. Suppose you agree an £8m pre-money and a £2m investment (£10m post, investor gets 20%). Now the investor requires a fresh 10% option pool. If it comes out of the pre-money, the founders' shares are diluted to make room for it, and the effective pre-money valuation drops. In practice the £8m headline pre-money now silently contains an £0.8m pool carved entirely from the founders — the real pre-money value attributable to founder shares is closer to £7.2m. The investor's 20% is untouched; the pool dilutes founders alone.
Negotiate the option pool as a post-money percentage, or size it to an actual hiring plan for the next 12–18 months rather than accepting a round "20%". Every point of pool taken from the pre-money is dilution borne entirely by founders and existing common holders — not by the incoming investor.
The no-shop clause is one of the few binding terms. It bars the founder from soliciting or negotiating with other investors for a defined period — typically 30 to 60 days — while the lead investor completes due diligence. It squarely favours the investor: it takes the company off the market at exactly the moment its negotiating leverage is highest, and it prevents a competitive process that might drive up the valuation. Founders should keep the exclusivity window short (30 days), ensure it starts on signing and has a hard end date, and confirm there is no automatic extension. Never sign a term sheet you are not genuinely ready to close — the no-shop is real even when the rest of the page is not.
Founder vesting means founders earn their already-owned shares back over time by continuing to work at the company. Mechanically this is reverse vesting: the founder owns the shares outright at closing, but the company has the right to buy them back (usually at nominal cost) on any shares that have not yet "vested" if the founder leaves. A standard schedule is 4 years with a 1-year cliff — nothing vests until you have stayed one year, then it accrues monthly.
This favours the investor and, importantly, the remaining co-founders: it ensures that a founder who quits after six months cannot walk away with a large slug of dead equity that demotivates everyone still building. It matters most in multi-founder teams, where the alternative is a departed co-founder holding a huge, unearned stake. Two terms to negotiate: vesting credit for time already served before the round, and acceleration — "single-trigger" (vesting accelerates on a sale) or the more common "double-trigger" (accelerates only if the company is sold and the founder is let go), which protects founders against being fired right before an acquisition to strip their unvested shares.
US deals are drafted around convertible preferred stock and the NVCA suite. UK deals use preference shares, the articles of association plus a shareholders' agreement, and the BVCA model documents; they also interact with tax-advantaged schemes (SEIS/EIS) whose rules restrict certain preferential terms — for example, redeemable shares and some preferences can disqualify EIS relief, which pushes UK seed rounds toward cleaner ordinary/preference structures. Always check EIS/SEIS compatibility before agreeing a preference or anti-dilution clause in a UK round.
Two things make a venture investor money: seeing the right companies, and choosing correctly among them. The industry romanticises the second — the flash of judgement, the contrarian bet — but the first is at least as decisive. You cannot back a company you never saw, and the very best deals are chronically oversubscribed, so being in the room early is not a nicety, it is the whole game. This chapter is about how deals reach an investor's desk, how many you must see to make a handful of investments, and where the fashionable promise of data-driven sourcing genuinely helps and where it quietly leaks.
Venture returns follow a power law: a tiny number of investments return the entire fund and then some, while most return little or nothing. Fred Wilson's much-argued "venture capital math problem" makes the shape concrete — a fund only works if a few holdings become enormous, because the median outcome is a write-off (AVC, 2009; and his candid 2020 revision after he underestimated exit sizes by an order of magnitude).
The power law has a brutal corollary for sourcing. If returns are concentrated in the top fraction of a percent of companies, then missing the winners costs you far more than picking a few losers. A fund with impeccable judgement but mediocre access will diligence carefully and invest wisely in the second tier — and still underperform, because the fund-returners never reached it. Access is a prerequisite for judgement to matter at all. This is why partners guard their networks jealously and why "we passed on that one" is a more forgivable sin than "we never saw it."
Judgement operates only on the set of companies you actually see. In a power-law asset class, the cost of a winner you never met dwarfs the cost of a loser you funded. Deal flow is the funnel that determines what judgement even gets to act on.
"Deal flow" is simply the stream of investable opportunities reaching a firm — its volume, quality, and velocity. It arrives through a handful of distinct channels, each with a different quality profile and a different cost to cultivate.
| Channel | What it is | Quality signal | Cost to build |
|---|---|---|---|
| Warm network | Founders, angels, lawyers, and past portfolio CEOs who refer people they know | High — a trusted referrer pre-filters | Years of relationships |
| Co-investor referrals | Other VCs sharing rounds they are leading or syndicating | High, but rarely "proprietary" — others see it too | Reciprocity and reputation |
| Accelerators / YC | Batch demo days and pre-demo access to cohorts | Curated but competitive and priced-up | Relationships with programmes |
| Outbound / proactive | Investors identifying targets and reaching out cold | Variable; you set the thesis | Analyst time, tooling |
| Scouts | Operators paid small cheques to spot deals in their niche | Extends reach into communities you lack | Programme + capital |
| Communities | Open-source, Slack/Discord, hackathons, university labs | Early but noisy | Presence and credibility |
| Inbound / data-driven | Applications, and signals surfaced by data pipelines | Highest volume, lowest average quality | Brand + engineering |
The academic picture confirms how relationship-heavy this is. In the large survey behind "How Do Venture Capitalists Make Decisions?" (Gompers, Gornall, Kaplan and Strebulaev), the majority of deals VCs pursue are generated through their professional networks and their own proactive sourcing rather than passively received — and, tellingly, the average firm considers roughly 100 opportunities for every one it closes. Relationships are not just pleasant; they are the primary intake valve.
UK/EU note: the structure is the same in London, Berlin, Paris, and Stockholm, but the intermediary layer differs. Tax-advantaged retail capital under SEIS/EIS in the UK feeds a dense angel network that seeds much early deal flow; the Y Combinator equivalent role is played by a spread of programmes (Entrepreneur First's talent-first model, Techstars, Seedcamp) rather than one dominant one. European seed rounds also lean more on angel syndicates and government co-investment vehicles, which changes who the trusted referrers are.
Proprietary deal flow is the claim that a firm systematically sees good companies before, or instead of, its competitors. It is the single most over-claimed phrase in venture. In practice, almost nothing is truly exclusive: a company raising a priced round talks to many investors, and a "proprietary" introduction usually just means you were early and trusted, not that you were alone.
The useful reframing is that proprietary access is not a static asset you own but a timing and relationship advantage you rent. Mark Suster's well-known argument that investors should "invest in lines, not dots" captures the mechanism: meet a founder long before they raise, watch their trajectory across several data points (the line), and you earn the right to lead when the round opens — not because the deal was hidden, but because you built conviction and trust while others were still seeing a single dot. That is as close to proprietary as early-stage venture gets.
"Proprietary deal flow" is mostly a myth if it means exclusivity. It is real if it means earned early access plus the trust to win a competitive round. The moat is the relationship and the timing, not secrecy.
For anyone building sourcing tools, this distinction is load-bearing. Software can reliably manufacture early awareness — surfacing a company months before it raises. It cannot, by itself, manufacture the trust that converts early awareness into an allocation. The tool gets you the dot early; the humans still have to draw the line.
Every firm, however it dresses it up, runs the same funnel. Opportunities enter at the top and are filtered at each stage:
Two forces shape the funnel. Each stage has a conversion rate (the fraction that advances), and the last stage has a win rate that is not fully in your control — issuing a term sheet does not guarantee the founder chooses you over a competing offer. The overall throughput is the product of all the stage conversions, which is why funnels are so unforgiving: a chain of individually reasonable percentages multiplies down to a fraction of a percent.
Take a concrete, realistic seed fund that intends to make 10 new investments per year. Work backwards from the target, applying plausible conversion rates at each stage.
| Stage | Count / year | Conversion to next | Note |
|---|---|---|---|
| Leads sourced/seen | 1,200 | 25% | ~100 a month across all channels |
| Screened (passed filter) | 300 | 40% | Right stage/sector, not obviously flawed |
| Met (first call) | 120 | 33% | ~10 first meetings a month |
| Diligenced (deep work) | 40 | 30% | Serious internal effort per company |
| Term sheets issued | 12 | 83% | You lose ~2 to competing offers |
| Closed | 10 | — | The target |
The end-to-end conversion is 10 ÷ 1,200 ≈ 0.83% — about 120 companies seen for every one closed, squarely in line with the ~100:1 the Gompers survey found. Notice where the funnel narrows hardest: the screen-to-meeting and diligence gates each cut the field by two-thirds or more, while the term-sheet-to-close win rate, though painful, moves the top-of-funnel requirement only modestly.
Now run the sensitivity that matters. Suppose your win rate collapses from 83% to 50% because you compete for hotter deals and lose more of them. To still close 10, you must issue 20 term sheets, which — holding every earlier conversion constant — requires roughly 2,000 leads instead of 1,200. Losing a third of your bids at the finish line inflates the entire top of the funnel by ~65%. Conversely, if a strong brand lifts your first-meeting conversion (founders arrive pre-sold and better-qualified) from 33% to 50%, you can hit 10 closes from about 800 leads. Small percentage shifts at the wide part of the funnel swing the required intake by hundreds of companies.
To make 10 seed investments you realistically need to see on the order of 1,000–2,000 companies a year. The exact number is governed less by how many you invite in and more by your conversion and win rates — which is why the cheapest way to grow is often to improve conversion, not volume.
There are two ways to make the funnel produce more good investments, and they trade off against each other. You can widen the top — see more companies — or you can improve conversion and selectivity at each gate. Naively, more volume looks strictly good. In practice it is not free: every additional lead consumes screening and meeting time, and beyond a point extra volume is dominated by lower-quality inbound that dilutes attention and pushes your team toward superficial screening. A funnel that doubles its leads but halves the care taken at each stage can easily close fewer good deals.
The healthiest firms therefore treat top-of-funnel volume and conversion quality as a balanced pair. They want enough volume to ensure the winners are statistically present in the pipeline, and enough conversion discipline that the winners are actually recognised and won. Data-driven sourcing is seductive precisely because it promises volume cheaply — but volume without a matching lift in screening throughput just relocates the bottleneck.
The most durable sourcing advantage is not a channel at all; it is reputation. A firm known for backing a category-defining company, or for treating founders well when things go wrong, changes its funnel in three compounding ways. First, it lifts the top: the best founders route to it unprompted, so inbound quality rises rather than falls. Second, it lifts early conversion: founders arrive already wanting the firm, so first meetings convert better. Third, and most valuable, it lifts the win rate: when a hot founder holds three term sheets, brand is the tie-breaker that lets you win at a fair price instead of overpaying.
This is why the funnel and the portfolio are coupled. A single fund-returning outcome is not just a financial win; it is a sourcing asset that improves conversion at every stage for a decade. Content, thesis-driven writing, and visible domain expertise (the model behind Fred Wilson's and Mark Suster's long-running blogs) work the same way at smaller scale: they make founders self-select toward you before any human in your firm has spent a minute. Reputation is the one input that improves volume, conversion, and win rate simultaneously — which is why partners treat founder references as sacred.
Data-driven sourcing means building a pipeline that ingests signals — hiring velocity, web traffic, GitHub and package-download activity, app rankings, patent filings, company-registry changes, founder pedigree — and surfaces companies matching a thesis before they hit the market. The infrastructure has become formidable: SignalFire's platform, for example, tracks millions of companies in real time across millions of data sources (California Management Review, 2023). For a team building sourcing tools, it is worth being precise about what this machinery does well and where it leaks.
What a data pipeline finds well:
Where it leaks:
Data-driven sourcing is best understood as an exploratory coverage layer bolted onto a relationship business — it fixes the top and middle of the funnel (coverage, timing, prioritisation) but not the bottom (winning the deal), and it is weakest precisely where the returns are richest: the pre-signal, pre-traction outlier. Build it to feed human conviction earlier, not to replace it.
The synthesis, then, is that sourcing is neither pure relationship craft nor pure engineering. The firms that win widen coverage with data, prioritise ruthlessly to protect human attention, and pour that attention into building trust early — drawing lines, not dots — so that when the round opens they are early, credible, and chosen. Access and judgement are not rivals; the funnel is where they meet.
Every cheque a venture investor writes is a bet on incomplete information about a future that mostly won't happen. Diligence is the discipline of buying down that uncertainty — not to zero, which is impossible, but to a level where a rational person can act. This chapter shows you what gets checked, how judgement is structured, where human cognition sabotages the process, and which parts of it a data pipeline can genuinely help with.
Due diligence is the investigation an investor runs between "interested" and "committed." Its purpose is not to prove the company will succeed — no amount of checking can do that at the seed stage — but to find the reasons not to invest, price the risks that remain, and confirm that the story the founder tells matches the evidence. A good process is deliberately adversarial toward its own enthusiasm. You already like the deal; diligence exists to give the deal a fair chance to fail your scrutiny before it fails your fund.
Two forces shape how much diligence actually happens. First, stage: at pre-seed there may be nothing to inspect but a founder and a prototype, so "diligence" is mostly judgement; by Series B there are years of financials, contracts, and customer data, and the process becomes forensic. Second, competition: in a hot round the investor who demands three weeks of data-room access loses to the one who wires in three days. Diligence is therefore always a negotiation between rigour and speed, and the best investors are explicit about which corners they are choosing to cut.
Formal diligence is conventionally split into four domains. Each answers a different question and, importantly, is usually run by different people — a mismatch between them is itself a signal.
| Domain | Core question | Representative checks |
|---|---|---|
| Commercial | Is there a real, growing market and a defensible position in it? | Market sizing (TAM/SAM/SOM), customer interviews, competitive landscape, pipeline quality, pricing power, churn and retention |
| Technical | Does the product work, and can it scale without collapsing? | Architecture review, code quality and security, IP ownership, technical debt, key-person dependency, roadmap feasibility |
| Financial | Do the numbers reconcile and do the unit economics hold? | Revenue recognition, MRR/ARR bridge, burn and runway, gross margin, CAC/LTV, cap table, prior financings |
| Legal | Does the company actually own what it claims, and is it clean? | Incorporation and share structure, founder vesting, IP assignment, employment and contractor agreements, key contracts, litigation, regulatory exposure, data protection (UK GDPR) |
Commercial and financial diligence overlap heavily and are often led by the investment team itself. Technical and legal diligence are frequently outsourced — a friendly CTO does the code review, a law firm does the legal — because they require expertise the deal team lacks and independence the deal team cannot supply. In the UK and EU, legal diligence carries extra weight around data protection (UK GDPR / the EU GDPR), IP chain-of-title (has every contractor and founder actually assigned their work to the company?), and EIS/SEIS eligibility, a UK tax-relief regime that many angels depend on and that certain deal structures can quietly disqualify.
The single most common fatal legal finding at early stage is broken IP assignment — a founder who built the prototype while employed elsewhere, or a freelance designer who never signed over the code. The company may not legally own its own product. This is cheap to check and catastrophic to miss.
Diligence gathers facts; the four classic evaluation lenses turn facts into a decision. They are team, market, product, and traction. The art is not in the lenses themselves but in how you weight them, and the weighting is a function of stage — because different stages have different things worth knowing.
The stage-dependence is the key insight for anyone building evaluation tools. At pre-seed and seed, there is almost no traction and no financials to analyse, so the decision leans overwhelmingly on team and market — you are underwriting people and a thesis. By Series A and beyond, traction and financials dominate, because now there is a real operating history and the question shifts from "could this work?" to "is this working, and will it keep compounding?" A scoring model that applies the same weights across stages will systematically mis-rank companies.
| Lens | Pre-seed / Seed | Series A | Series B+ |
|---|---|---|---|
| Team | Very high | High | Medium |
| Market | Very high | High | Medium |
| Product | Medium | High | Medium |
| Traction / Financials | Low (little exists) | High | Very high |
Because team is the heaviest lens and the hardest to read from documents, experienced investors treat reference calls as core diligence, not a formality. There are two kinds. On-list references are the people the founder gives you — useful, but curated to be positive. Back-channel references are the ones you find yourself: a former colleague, an ex-investor, a customer who churned. These are where the truth lives, precisely because the founder didn't choose them.
The technique that separates good reference work from theatre is asking questions that make it socially easy to say something negative. "Was she good?" invites a reflexive yes. Better prompts, drawn from how top operators run executive references: "Where did this person spike, and where did they need support?"; "Would you jump at the chance to work with them again — and if there was any hesitation in that answer, what was it?"; and the pattern-probe, "I've heard from a couple of people that X — what's your read on that?" (First Round Review). The same methods VCs use to reference founders are the ones founders use to reference VCs — the relationship is genuinely two-way, and the best founders back-channel their investors just as hard.
Back-channelling has an ethics and a discretion cost. Calling a founder's current employer or largest customer without permission can leak the fundraise, damage the founder, and poison the relationship before you've invested. Agree the boundaries of off-list contact with the founder up front, and treat everything you hear as one data point, not a verdict — disgruntled ex-colleagues distort as reliably as loyal ones flatter.
The data room is the shared repository (today usually a permissioned Google Drive or a purpose-built tool) where the company assembles the evidence. A well-organised data room is itself a signal: founders who can produce clean, complete, well-labelled documents quickly tend to run tight companies. A chaotic or slow-to-materialise data room predicts chaotic operations.
A standard early-stage request list:
What you ask for should mirror what a well-prepared founder already has. The canonical Sequoia framing of a business plan — purpose, problem, solution, why now, market, competition, business model, team, financials, vision — doubles as a diligence outline, because it lists exactly the claims you then go and verify (Sequoia). For the founder's side of this dance and how the process feels from their seat, YC's guide to seed fundraising is the standard reference.
When traction exists, the most decision-relevant analysis is not the headline growth number — which is easy to juice with marketing spend — but the quality of that growth. Two tools carry most of the weight.
Cohort analysis groups customers by when they joined and tracks each group over time. It answers the question a topline chart hides: are customers staying? A company can post rising revenue while every cohort quietly bleeds out, kept afloat only by ever-larger acquisition. Flat or upward-sloping retention curves — especially net revenue retention above 100%, where existing customers spend more over time than churned ones take away — are among the strongest positive signals in all of diligence, because they mean the business compounds rather than leaks (David Skok, SaaS Metrics 2.0).
Unit economics ask whether one customer is profitable over their lifetime. The two workhorse figures are CAC (fully-loaded cost to acquire a customer) and LTV (gross-margin profit that customer generates before churning). The rough conventions: LTV:CAC of roughly 3:1 or better, and CAC payback under about 12 months. Below those thresholds, growth destroys cash; above them, every pound of spend is an investment rather than a subsidy. Beware the common manipulations: CAC that excludes salaries, LTV that assumes optimistic lifetimes, and blended CAC that hides expensive paid channels behind cheap organic ones.
The uncomfortable truth is that most venture decisions are made by human beings on thin data under time pressure — exactly the conditions where cognitive bias flourishes. Knowing the named failure modes is the first line of defence.
| Bias | How it shows up in VC |
|---|---|
| Pattern-matching | "This founder reminds me of [famous successful one]." Efficient heuristic, but it overfits to superficial traits (school, demeanour, prior employer) and entrenches whoever succeeded before. |
| Halo effect | One impressive attribute — a Stanford PhD, a hot logo customer — bleeds a positive glow over everything else, so unrelated weaknesses go unexamined. |
| Confirmation bias | Once you like a deal, diligence becomes a search for supporting evidence rather than disconfirming evidence. You call the on-list references and skip the awkward back-channel. |
| Herding | The presence of a respected co-investor substitutes for independent analysis. "If Sequoia's in, it must be good" — which is how whole rounds get mispriced at once. |
| Affinity bias | Preference for founders who resemble the investor — same background, gender, network. This is both an accuracy problem and a structural driver of who gets funded. |
Structured and data-driven processes exist to counteract these. Standardised scorecards force every lens to be considered rather than letting a halo carry the decision. Pre-committed thesis memos, written before diligence, make confirmation bias visible when reality diverges. Blind or structured first-screens reduce affinity bias. Deliberately seeking disconfirming evidence — the "pre-mortem," where the team writes the story of how this investment failed — is a direct antidote to motivated reasoning.
Structure is not a cure, and can become camouflage. A scorecard whose weights were chosen to justify a decision already made launders bias into apparent objectivity. Herding survives quantification — everyone's model uses the same inputs. And a data pipeline trained on who got funded and succeeded historically will faithfully reproduce the field's affinity and pattern-matching biases, now with a veneer of algorithmic neutrality. Automating a biased judgement scales the bias; it does not remove it.
A usable seed-stage checklist, with an honest column on what a data pipeline can and cannot do — the distinction that matters most if you are building evaluation tools.
| Item | What you're checking | Automatability |
|---|---|---|
| Cap table clean & founders vested | Ownership sensible, no dead equity | Well — structured data, rule-checkable |
| IP fully assigned to company | Company owns its product | Poorly — needs document reading + judgement |
| Market size & "why now" | Prize is big; timing is real | Partly — can gather comps; the "why now" is a thesis, not a number |
| Growth rate (MoM/YoY) | Momentum | Well — pure calculation |
| Cohort retention / NRR | Growth compounds, doesn't leak | Well — given clean data |
| Unit economics (LTV:CAC, payback) | Growth creates value | Well to compute, poorly to trust — inputs are easily gamed |
| Customer concentration & pipeline | Durability of revenue | Partly — countable; quality needs calls |
| On-list + back-channel references | Founder integrity & capability | Can't touch — human, contextual, deniable |
| Founder-market fit & resilience | Will they endure and adapt? | Can't touch — the core judgement call |
| Legal/regulatory exposure | Hidden liabilities | Partly — flags searchable, materiality needs a lawyer |
The pattern is consistent and worth internalising. Data pipelines automate the countable and the structured extremely well — growth, retention, cap-table hygiene, screening thousands of companies to surface candidates. They handle document-heavy checks poorly, because materiality is contextual — a pipeline can flag a missing IP assignment but can't tell you whether it's fatal or trivial. And they cannot touch the highest-weighted, stage-critical judgements at all: founder integrity, resilience, market timing, the truth that only a reluctant back-channel reference will tell you. The most valuable role for automation is therefore not to make the decision but to clear the mechanical work so the human can spend their scarce judgement where it actually decides the outcome — and, crucially, to enforce the discipline (every box considered, disconfirming evidence sought) that bias erodes. As Mark Suster's much-cited framing has it, investors ultimately back lines, not dots — the trajectory of a person over time, which no snapshot, however data-rich, can fully capture.
At the earliest stages a startup's "valuation" is not the output of a spreadsheet — it is a number two parties agree on because it produces an ownership split both can live with. There is no cash flow to discount, no profit to multiply, often no revenue at all. This chapter builds valuation from first principles: the mechanics that are pure arithmetic (pre-money, post-money, ownership), the judgement that is pure negotiation (what the number should be), and the disciplined middle ground the venture-capital method provides. By the end you will be able to take a headline like "raised £2m at £10m post" and reconstruct exactly what the investor bought, what they must believe, and what has to happen at exit for the deal to work.
Start by separating two things that beginners constantly conflate. The mechanics of a priced equity round are simple algebra that never varies. The valuation itself — the figure plugged into that algebra — is a negotiated price, set far more by supply, demand, and story than by any formula. Most of the confusion in startup finance comes from treating the second like the first: reaching for a calculation to justify a number that was really decided by how many investors wanted in.
The mechanical layer rests on two definitions:
post = pre + investment.Ownership follows directly. The investor's stake is the check divided by the post-money value:
investor ownership % = investment ÷ post-moneypre-money = post-money − investmentThat single fraction — check over post — is the most important equation in early-stage finance, because it means the post-money figure and the check jointly determine ownership. You cannot change one without changing the split unless you change the other. This is why sophisticated investors negotiate in terms of percentage owned, while founders often anchor on the headline pre-money — they are looking at opposite ends of the same equation.
Take the headline "raised £2m at £10m post." The post-money is £10m and the check is £2m, so the investor owns £2m ÷ £10m = 20%. The pre-money was £10m − £2m = £8m. The founders, who owned 100% of an £8m company the instant before signing, now own 80% of a £10m company — £8m of value on paper, unchanged, exactly as intended: the investor's cash sits on the balance sheet and belongs to everyone pro-rata. Notice how the two framings feel different but say the same thing. "I'll invest £2m at £8m pre" and "I need 20% for my £2m" are identical deals. If the founder pushes the pre to £9m, the same £2m now buys £2m ÷ £11m = 18.2% — the founder gains 1.8 points of ownership, and the investor's target erodes. Every pound of pre-money is a direct transfer of ownership from investor to founder.
Here is the counter-intuitive part. A great many venture firms do not start from "what is this company worth?" at all. They start from "what percentage do I need to own?" and back into the valuation. This is ownership-target-driven pricing, and it exists because of how funds make money.
A VC fund returns capital through a handful of enormous winners; the mathematics of the portfolio (covered in the fund-economics chapter) demand that each winning position be large enough to move the whole fund. If a fund needs a winner to return, say, £50m to matter, and it can only own a stake that gets diluted over time, it must secure a meaningful slice — commonly 10–25% — at entry. Below some floor, the position simply cannot pay back the fund even if the company succeeds, so the firm passes regardless of price. As Mark Suster puts it, VCs "calculate valuation differently from founders" precisely because they reason backwards from ownership and fund returns rather than forwards from the company's merits.
Once the target percentage is fixed, the valuation is a residual:
post-money = investment ÷ target ownership %pre-money = post-money − investmentAn investor who must own 20% and is writing a £2m check is defining a £10m post-money and an £8m pre-money — not discovering it. Flip the causality and the whole negotiation makes sense: the check size and the ownership target set the valuation, not the other way round. This is also why round sizes and valuations move together. A founder who wants to raise more money without giving up more ownership must argue for a higher valuation; a founder who accepts a lower valuation must either sell more of the company or raise less.
Ownership targets tell an investor how much to buy but not what price is defensible. The bridge is the venture-capital method, formalised in teaching by Aswath Damodaran's chapter on VC valuation. Instead of discounting cash flows the company doesn't yet have, it reasons from a single imagined future — the exit — back to today. The logic runs in four steps:
ownership at exit = (investment × target multiple) ÷ exit value.ownership today = ownership at exit ÷ expected retention ratio.The implied post-money then falls straight out: post-money = investment ÷ ownership today. The method's virtue is that it makes every assumption explicit and negotiable. Its vice is that it is only as good as the exit guess, which at seed is close to a fiction — hence the enormous target multiples that absorb the uncertainty.
The full round, end to end. An investor puts £2m into our company at £10m post-money, taking 20% today. What must they believe?
Return target. A 10x on £2m means the position must be worth £20m at exit. Over a 6-year hold that 10x is an annual return of 10^(1/6) − 1 ≈ 47% IRR — a useful sanity check on how demanding "10x" really is.
Dilution to exit. Suppose two more priced rounds each sell roughly 20% of the company. The seed investor's stake is multiplied by 0.8 × 0.8 ≈ 0.64, and option-pool top-ups drag it a little lower — call the retention ratio 0.60. Their 20% today becomes about 20% × 0.60 = 12% at exit.
Required exit value. For a 12% stake to be worth £20m, the company must exit at £20m ÷ 0.12 ≈ £167m. So the entire deal rests on a single implicit claim: this £10m-post company can plausibly be worth ~£167m in six years.
Now run it the other way, VC-method style. Say the honest exit view is a sale at £150m (e.g. £30m of year-6 revenue at a 5x revenue multiple). Required ownership at exit = £20m ÷ £150m = 13.3%. Grossed up for 0.60 retention, ownership today = 13.3% ÷ 0.60 = 22.2%. The implied post-money = £2m ÷ 0.222 = £9.0m, i.e. a £7m pre-money — a full million below the £8m pre the founder wanted. That £1m gap is the negotiation. The founder argues the exit is bigger than £150m or dilution is gentler than 40%; the investor argues the opposite. Nobody is wrong — they are pricing different futures, and the deal closes when the two stories overlap.
The instinct of a finance-trained reader is to reach for a discounted cash flow (DCF) or a revenue multiple. At the earliest stages both largely break, and it is worth understanding precisely why rather than taking it on faith.
A DCF values a company as the present value of its future cash flows. For a pre-revenue seed company, essentially all of that value sits in a terminal value many years out, resting on growth and margin assumptions no one can defend. Discounting a number you invented by a rate you also invented produces false precision, not insight — Damodaran's paper on valuing young companies works through exactly how fragile these inputs are and how the value swings wildly on tiny assumption changes. Revenue multiples fail for a blunter reason: a multiple needs a denominator. Zero or trivial revenue makes "5x revenue" meaningless, and near-zero revenue makes it absurdly sensitive — a company doubling from £100k to £200k of ARR would "double in value" on a naive multiple, which no investor believes.
These methods start to apply as the denominator stabilises. By Series A–B, with a few million in recurring revenue and a visible growth rate, a forward revenue multiple becomes a genuine market signal; by later growth rounds and certainly at exit, DCFs and earnings multiples do real work. The transition is gradual: the more predictable the cash flows, the more the calculation earns its place and the less the price is pure negotiation. Early-stage valuation is negotiation because the calculation cannot yet bear weight — not because founders and investors are being unrigorous.
In the vacuum left by DCF, the working tool is comparables: what did similar companies, at a similar stage, in a similar sector and geography, raise at recently? Benchmarks anchor a negotiation to market reality and stop both sides drifting into fantasy. But they are a starting point, not an answer, for three reasons. First, "similar" is doing heavy lifting — team, traction, and market size vary enough that two seed rounds can justifiably price 3x apart. Second, benchmarks are lagging and cyclical: they reflect the market that closed six months ago, and valuations compress or inflate fast. Third, they are reflexive — everyone quoting the same comps inflates the comps. Treat published medians as a sanity band, not a target. Indicative UK/EU ranges (label: these move with the cycle and should be checked against current data):
| Stage (UK/EU) | Typical round | Typical post-money | Typical stake sold |
|---|---|---|---|
| Pre-seed | £250k–£1m | £2m–£6m | 10–20% |
| Seed | £1m–£3m | £6m–£15m | 15–25% |
| Series A | £4m–£12m | £20m–£60m | 15–25% |
| Series B | £12m–£30m | £60m–£150m | 15–20% |
Note how the percentage sold stays remarkably stable across stages even as absolute numbers explode. That stability is the ownership-target logic operating market-wide: investors keep buying similar slices; valuations rise to accommodate larger checks.
The headline valuation hides a lever that quietly changes the real price: the employee option pool. Investors almost always require a pool of unissued shares (commonly 10–15%) to hire and retain talent — and they typically insist it be created before their investment, carved out of the pre-money. That placement matters enormously. A pool created pre-money dilutes only the founders; a pool created post-money would dilute everyone. Fred Wilson is blunt that this is "just another way to lower the price" — a £8m pre-money with a fresh 15% pool inside it is really worth about £6.8m to the founders, because £1.2m of that "value" is shares they are giving away to future employees.
Return to the £2m at £8m pre / £10m post deal, and now require a 15% post-financing option pool carved from the pre-money. On a fully-diluted basis the pool takes 15% of the company off the top; the investor still takes 20%; the founders keep the remaining 65% — not the 80% the headline implied. The investor's stake is unchanged, so the entire 15% comes out of the founders' hide. To hold the founders at 80%−(a pool from post-money would be gentler), they must negotiate either a smaller pool or a higher pre-money. The lesson: the pool's size and its placement are part of the price, and belong in the same conversation as the valuation, not a footnote after it.
The percentage you own today is not the percentage you own when the money is made. Every subsequent round issues new shares and shrinks existing holders — this is dilution, and it is inevitable for any company that keeps raising. Fred Wilson's walk-through of dilution shows founders falling from 100% to roughly 42% across a seed and two venture rounds, with the earliest shareholders diluted hardest. The consolation, correctly stated, is that a smaller slice of a much larger pie can be worth far more — which is exactly why the VC method grosses up for the retention ratio. An investor who ignores dilution and prices to their day-one stake will systematically overpay, because the stake that actually cashes out at exit is materially smaller than the one they bought.
All the arithmetic above establishes a defensible range. What picks the actual number inside that range is competition for the deal. Early-stage valuation is a market: a founder with three term sheets prices at the top of the band; a founder with one, running low on cash, prices at the bottom or worse. The scarce, expensive resource is investor conviction, and it is allocated by demand. This is why the same company can be worth £6m pre with one interested party and £12m pre when a hot round is oversubscribed — nothing about the business changed, only the auction. It is also why "hot" sectors carry inflated valuations that later revert: money floods in, demand spikes, prices overshoot the fundamentals the VC method would support, and the correction arrives a round or two later. Damodaran's ongoing commentary at Musings on Markets is a good antidote to the narrative excess that drives these swings. The practical upshot for a founder is to engineer competitive tension and time the raise into strength; for an investor, it is to know your walk-away price from the return math and not let auction momentum push you past it.
Valuation at the earliest stages is a negotiation disciplined by arithmetic, not an arithmetic dressed up as a negotiation. The mechanics — check over post equals ownership — are fixed and unforgiving. The number fed into them is chosen: anchored by ownership targets and the return math of the VC method, bounded by comparables, quietly adjusted by the option pool and future dilution, and finally set by supply and demand on the day. Master the mechanics so you are never confused about what a deal actually does; treat the number itself with humility, because until the cash flows arrive, it is a shared bet on the future — and both sides know it.
A venture fund's spreadsheet can show a portfolio "worth" ten times what it cost, and every pound of that can still be worthless. A markup is an opinion; an exit is a fact. This chapter is about the moment opinion becomes fact — when a private company's shares turn into cash or freely tradable stock, and the carefully layered rights on the cap table decide who actually walks away with what.
Venture capital is an illiquid asset class. When a fund invests, it buys preferred shares in a private company — shares with no public market, no bid, no ask. The fund can revalue those shares upward every time a new round prices higher, and it reports this as unrealised value. Limited partners (the pension funds, endowments and family offices whose money the fund invests) see two numbers: TVPI (total value to paid-in — realised plus paper value, divided by capital called) and DPI (distributions to paid-in — actual cash returned). TVPI is the promise; DPI is the delivery. A fund can post a 4x TVPI for years and, if its winners never exit, return a 0.8x DPI and lose its investors money.
An exit (also called a "liquidity event") is any transaction that converts illiquid preferred shares into distributable value: a trade sale, an initial public offering, or a secondary sale of the shares themselves. Only an exit produces DPI. This is why the entire venture model — the willingness to lose money on most investments — hangs on the few that exit large. Correlation Ventures' analysis of ~21,000 US financings and later AngelList's own dataset both confirm a brutal power-law distribution: a heavy right tail where a single 50x or 100x exit outweighs dozens of write-offs. No exit, no tail, no fund.
Most exits are acquisitions, not flotations. In the US the NVCA Yearbook shows M&A dwarfing IPOs by count year after year; the UK/EU picture is more extreme still, because European public markets absorb far fewer venture-scale tech listings. If a venture-backed company achieves any meaningful exit at all, the base-rate expectation is that it gets bought.
Acquirers come in two flavours, and which one you face changes the price logic:
Two structures recur at the small end. An acquihire is an acquisition dressed as a talent purchase: the buyer wants the engineering team, the product gets shut down, and the price is often close to (or below) the money invested. For founders this is a soft landing; for investors holding a liquidation preference it can mean getting capital back with little upside. An earnout defers part of the price, paying it only if the acquired business hits agreed targets (revenue, retention, product milestones) over one to three years post-close. Earnouts bridge disagreement about value — the seller believes the growth story, the buyer wants proof — but they shift risk onto the sellers, who no longer control the business delivering their payout. Always separate the headline price from the guaranteed price: a "£100m acquisition" with £40m in earnout is an £60m certain deal.
An initial public offering sells shares to public-market investors and lists them on an exchange (Nasdaq or NYSE in the US; the London Stock Exchange or Euronext in the UK/EU). IPOs generate the headlines and the biggest single outcomes, but they are rare — a small fraction of venture exits — and available only to companies large and predictable enough to withstand quarterly public scrutiny.
Mechanically, at a high level: the company files a registration document (an S-1 with the SEC in the US; a prospectus approved by the FCA for a Main Market listing in the UK), underwriting banks run a roadshow to gauge institutional demand, a price is set the night before trading, and the shares begin trading the next morning. Two features matter for venture investors. First, an IPO is usually not an instant cash exit: at listing, all preferred shares automatically convert to common, and the fund now holds public — but restricted — stock. Second, the lock-up: underwriters typically bar insiders (founders, employees, VCs) from selling for 90–180 days after listing, to prevent a flood of supply crashing the price. The real exit for a fund is the post-lock-up distribution or sale, at whatever price the market has settled to by then — which can be well below the IPO price.
Companies now stay private far longer than they did in the 2000s — often 10–12 years to exit. That is a problem for early backers and employees whose shares are theoretically valuable but practically frozen. Secondaries solve it by selling the existing shares (not new shares the company issues) to another buyer, giving partial liquidity while the company remains private.
Fred Wilson's widely cited guidance on founder liquidity captures the norms: take some risk off the table only once the company has "sustainable and lasting enterprise value," and sell no more than ~10% of a position in one transaction, so incentives stay aligned. For a fund, selling a secondary stake means locking in a real DPI early — sometimes wise, because a 5x realised beats a 15x that never materialises.
When cash from an exit is distributed, it does not split by ownership percentage. It flows through a waterfall defined by the rights attached to each share class. Preferred shareholders hold a liquidation preference: the right to get paid before common shareholders (founders and employees). Two dials define it.
The multiple is how many times their money preferred get back first. The market standard is 1x — Cooley's Q4 2024 data shows 96% of US deals at 1x. Multiples above 1x reappear in hard markets and punish common brutally.
The participation dial is the one founders underestimate:
Preferences also stack by seniority. Later rounds are often senior to earlier ones (last money in, first money out) or rank pari passu (equal). In a shortfall, senior money is paid first and juniors may get nothing.
Reconstruct a mediocre exit. A UK startup raises three rounds, all with a 1x preference, and is acquired for £60m — a 2x on the £30m invested, respectable-looking but far below the last round's implied value. The as-converted cap table:
| Holder | Invested (1x pref) | As-converted ownership |
|---|---|---|
| Founders + employees (common) | — | 51% |
| Seed | £2m | 8% |
| Series A | £8m | 16% |
| Series B (most senior) | £20m | 25% |
Scenario 1 — 1x non-participating (market standard). Each investor takes the greater of pref or converting. At £60m, a pro-rata share is worth: Seed 8% = £4.8m (beats £2m pref → convert); Series A 16% = £9.6m (beats £8m → convert); Series B 25% = £15m (below its £20m pref → take the pref). So Series B takes £20m off the top. The remaining £40m is split among the converting holders (Seed 8%, A 16%, Common 51% — together 75% of shares): Common gets 51/75 × £40m = £27.2m; Series A £8.53m; Seed £4.27m. Investors collect £32.8m; founders/employees keep £27.2m.
Scenario 2 — 1x participating (double-dip). Every preferred takes its pref first: £2m + £8m + £20m = £30m off the top. The remaining £30m is then split by full ownership, preferred included: Common 51% = £15.3m, Series B +£7.5m, Series A +£4.8m, Seed +£2.4m. Final tally: Series B £27.5m, Series A £12.8m, Seed £4.4m, and Common just £15.3m.
Same company, same £60m, same 1x — but flipping one word ("participating") moves £11.9m from the founders and team to the investors. Now push the price down to £30m (a 1x on capital): the £30m in prefs consumes the entire proceeds, Series B and A take their full prefs, and the common column — founders and every employee's options — reads £0. A "£30m sale" can be a total wipeout for the people who built the company. This is the mechanism behind headlines where a company "sells for millions" yet the founders get nothing.
The non-participating choice — take the preference, or convert to common — has a clean break-even. An investor converts when their ownership percentage of the total proceeds exceeds their preference. In the example, Series B's break-even is £20m ÷ 25% = £80m: below an £80m exit it clings to its £20m pref; above it, converting to common is worth more and it drops the preference. The preference is downside insurance that the investor voluntarily discards once the upside is big enough. This is exactly why, in genuine home-run exits, preferences become irrelevant — everyone converts to common and proceeds split by ownership, as intuition would suggest. Preferences only bite in the mediocre and bad outcomes, which is precisely where most of the distribution lives.
Cash returned is only half the story; when it comes back drives the fund's internal rate of return. A fund's headline multiple is MOIC (multiple on invested capital); its annualised return is approximately IRR = MOIC^(1/years) − 1. Time compounds relentlessly against you.
| Outcome | Held 5 years | Held 10 years |
|---|---|---|
| 10x MOIC | ~58.5% IRR | ~25.9% IRR |
| 3x MOIC | ~24.6% IRR | ~11.6% IRR |
The same 10x exit is a spectacular 58% IRR if it lands in year five and a merely good 26% if the company drifts to year ten. This is the quiet cost of companies staying private longer, and a core reason funds increasingly welcome secondaries: a 5x realised in year six can beat a 10x that arrives in year twelve. For UK/EU angels there is a tax overlay too — EIS/SEIS relief requires holding qualifying shares for at least three years, so the exit clock interacts with after-tax returns, not just gross IRR. Fluency in exits means holding all three numbers at once: how much (MOIC), how it splits (the waterfall), and how long it took (IRR).
Every venture fund you have ever heard of is itself funded. Behind the general partners who write cheques to founders sits a second, quieter tier of capital: the limited partners, or LPs — the pension savers, universities, insurers and governments whose money actually flows into startups. Understanding VC without understanding LPs is like studying a river and ignoring the rainfall. This chapter maps who the funders' funders are, what they want, and what a new GP must grasp about the money sitting above them.
A venture capital fund is almost always a limited partnership — a legal vehicle with two kinds of member. The general partner (GP) runs the fund: sources deals, makes investments, sits on boards, and eventually returns money. The limited partners (LPs) supply the vast majority of the capital and, in exchange for limited liability, stay passive — they cannot direct individual investments without risking their legal protection. A typical fund is 98–99% LP money and 1–2% GP money (the "GP commit", discussed below).
The economics that bind them are the 2-and-20 model: roughly a 2% annual management fee on committed capital to run the firm, and 20% carried interest — the GP's share of profits — usually payable only after LPs get their money back plus a hurdle rate (commonly 8%). This structure exists because LPs are delegating: they lack the time, networks, or specialist skill to pick early-stage companies themselves, so they pay professionals to do it and align incentives through carry.
VC is a delegation chain. Savers give money to institutions; institutions (LPs) give money to funds; funds (GPs) give money to founders. Each layer takes a fee and adds selection. A new GP is not raising from "investors" in the abstract — they are raising from institutions that answer to their own beneficiaries and boards.
"LP" is a single label for a very diverse set of institutions, each with distinct motivations, time horizons, and constraints. The main categories:
| LP type | Core motivation | Typical constraints |
|---|---|---|
| Public pension funds | Fund retirement liabilities; beat actuarial return target (often ~6–7%) | Fiduciary duty, liquidity needs, illiquid-asset caps, political scrutiny, large minimum cheques |
| Corporate pensions | Meet defined-benefit obligations at low risk | Often de-risking/closing to new members; shrinking illiquid appetite |
| Endowments | Perpetual real growth of the corpus | Annual spend rate (~4–5%); reputational sensitivity |
| Foundations | Grow corpus while meeting minimum charitable spend | Mission alignment; spend obligations |
| Funds-of-funds | Diversified access; sell manager selection as a service | Double fee layer; must beat direct-investing net of costs |
| Family offices | Wealth preservation + growth; access, relationships, occasionally passion | Idiosyncratic; less process, more personal conviction |
| Sovereign wealth funds | Diversify national wealth; long-term returns; sometimes strategic aims | Very large minimums; governance and geopolitical sensitivity |
| Insurers | Match long-dated liabilities; yield | Solvency II capital charges on illiquids; regulatory reporting |
| Corporates | Strategic insight, sector access, optional returns | Shifting corporate priorities; shorter patience |
| Government programmes | Catalyse a domestic VC ecosystem; crowd in private money | Policy mandates, geographic/stage restrictions, state-aid rules |
It is tempting to say "returns", but that is only the headline. LPs weigh several things at once, and a GP who understands the full picture raises money faster.
Returns relative to a benchmark. An LP does not judge a venture fund in isolation. It asks: did this beat what I could have earned in public equities, and did it beat other VC funds of the same vintage year (the year the fund started deploying)? The standard yardstick is the public market equivalent (PME) — a comparison of fund cash flows against a public index like the S&P 500 or MSCI. If a VC fund can't beat public markets net of its hefty fees and illiquidity, the LP would rather buy the index. Benchmark data comes from providers such as Cambridge Associates, whose private-investment benchmarks are an industry reference.
The right performance metrics. LPs track several, because each hides something:
Diversification and access. Venture is a slice of an LP's total portfolio (often a low-single-digit to low-double-digit percentage). They want exposure to the asset class's upside without betting the fund on it. Crucially, in VC the returns are extraordinarily concentrated in a handful of top funds — access to those managers is itself the prize. Many LPs invest in a mediocre fund mainly to earn the right to re-up into the next, better one.
Constraints that shape behaviour. LPs face liquidity needs (pensions must pay out), regulatory capital rules (Solvency II for insurers), fiduciary and ESG obligations, and the denominator effect: when public markets fall, the fixed-dollar value of private holdings becomes a larger share of a shrinking portfolio, pushing LPs over their target allocation and forcing them to slow new commitments. This is why VC fundraising freezes when public markets crash — not because venture got worse, but because LP portfolios mechanically ran out of room.
A GP is not selling a return number. They are selling a credible story about future DPI, access to deals the LP can't reach alone, and a manager who will still be raising funds in a decade. LPs underwrite people and process as much as past performance.
Fundraising is a months-to-years process, not an event. The mechanics every new GP must know:
The pitch and diligence. A GP markets to prospective LPs with a deck and a data room. LPs run deep due diligence: track record (ideally attributed to named partners, not just the firm), team stability, sourcing edge, references from founders and co-investors, and operational/back-office competence. First-time funds face a chicken-and-egg problem — no track record means anchor LPs (often family offices, FoFs, or government programmes) take outsized importance in getting to a first close.
The LPA. The Limited Partnership Agreement is the contract governing the fund. It sets the fund term (classically "10+2" — ten years plus two one-year extensions), the investment period (typically the first ~5 years, during which new investments are made), management fees, carried interest and hurdle, the waterfall (the order in which cash is distributed), key-person clauses (if named partners leave, investing pauses), no-fault divorce provisions, recycling rules, and reporting obligations. The Institutional Limited Partners Association (ILPA) publishes model LPA templates and principles that have become the reference point for fair terms.
Capital calls, not lump sums. LPs commit capital but don't hand it over up front. The GP issues capital calls (drawdowns) over the investment period as deals arise, usually with ~10 business days' notice. Failing to meet a call triggers punitive default provisions. This is why LPs manage liquidity carefully — a commitment is a multi-year obligation, not a single payment.
Side letters. Large or early LPs negotiate side letters — bilateral agreements granting bespoke terms on top of the LPA: fee discounts, co-investment rights, enhanced reporting, ESG or excuse rights (opting out of certain investments), or most-favoured-nation (MFN) clauses that let them claim any better term granted to another LP. Side letters are a normal part of the negotiation but add administrative complexity a new GP must be ready to manage.
The GP commit. LPs insist the GP invests its own money — the GP commitment, historically ~1% of the fund but increasingly larger. This is "skin in the game": it ensures the GP loses alongside LPs if the fund underperforms, aligning incentives. For emerging managers without deep pockets, funding a meaningful GP commit is a real practical hurdle.
Ongoing reporting. Once closed, the GP reports quarterly (financials, valuations, portfolio updates) and holds an annual meeting and often an LP Advisory Committee (LPAC) — a subset of LPs who vote on conflicts of interest and valuation matters. Transparency is not optional: sloppy reporting is a common reason LPs decline to re-up.
A fund-of-funds is both an LP (to the underlying VC funds) and a fund (to its own investors). It exists to solve two problems for smaller or less specialised LPs: diversification (one commitment buys exposure to 10–20 underlying funds and hundreds of companies) and access (an experienced FoF may get into oversubscribed top-tier funds that a small pension never could). The cost is a second fee layer — you pay the FoF's fees on top of the underlying funds' 2-and-20, which drags net returns. FoFs justify this only if their manager selection genuinely outperforms. For a new GP, FoFs and government programmes are often the most reachable institutional LPs, because backing emerging managers is part of their stated mandate.
The European LP base differs from the US in structure and scale, and the differences matter enormously for a UK/EU GP.
The industry bodies. The trade association for UK private equity and venture is the BVCA — now rebranded UK Private Capital — which publishes performance data, standardised reporting, and policy advocacy. At the pan-European level, the equivalent is Invest Europe, whose investor guidelines and professional standards are widely adopted across the continent. These bodies matter to a GP because their templates and codes shape what LPs expect.
The cornerstone government LP. The single most important structural fact in UK venture is the British Business Bank and its commercial arm, British Patient Capital. Because the UK's private pension and endowment base historically allocated far less to venture than US institutions, government capital fills the gap — the British Business Bank is a cornerstone LP in a large share of UK VC funds, deliberately "crowding in" private money by committing early. The EU analogue is the European Investment Fund (EIF), which for decades has been the anchor LP of European venture. A UK/EU GP raising a first fund will very likely have these institutions high on their target list.
Pension pooling and the "Fit for the Future" reforms. UK local authority pensions are being consolidated into a small number of large LGPS asset pools under a government programme to build scale and unlock investment in productive assets, including venture and growth capital. The GOV.UK asset-pooling guidance sets the framework. Alongside it, the Mansion House agreements have seen large UK defined-contribution pension providers pledge to raise allocations to private markets — a deliberate policy push to close the gap with the US, where pensions and endowments have long been venture's backbone. For GPs, this signals a slowly deepening domestic institutional LP pool over the coming decade.
The angel and micro-fund layer: SEIS/EIS. Uniquely, the UK's earliest-stage capital is shaped by tax policy rather than institutions. The Seed Enterprise Investment Scheme (SEIS) gives individual investors 50% income-tax relief on investments (up to £200,000 per year) into very early startups (companies with gross assets under £350,000 and fewer than 25 employees, raising up to £250,000). The Enterprise Investment Scheme (EIS) extends 30% relief to larger raises, with companies able to raise up to £10 million a year and £24 million over their lifetime (higher for knowledge-intensive companies), plus capital-gains and loss reliefs. The effect is structural: SEIS/EIS turns thousands of UK high-earners and angels into a de facto seed-stage LP base, and underpins the economics of many UK micro-VC and syndicate funds. There is no direct US equivalent at this scale — it is one of the defining features of the UK early-stage market.
The US venture base rests on deep private institutional capital (endowments, pensions, family wealth). The UK/EU base leans far more heavily on government cornerstone LPs (British Business Bank, EIF) and, at the seed end, on tax-incentivised individual investors (SEIS/EIS). A GP fundraising in Europe should expect a different, more policy-shaped LP mix — and target it accordingly.
The money above you has its own bosses, its own clocks, and its own rules. Practical implications: