What the document has to contain, where the benchmark data actually comes from, and how to hold the line when a VC pushes back.
A startup valuation report for investors shows how you got to a number. What the number is takes one line; the rest of the document is the reasoning behind it, carried by six sections.
We have produced valuation reports for 160,000+ companies (Equidam internal data) across 90+ countries and 600+ industries (Equidam data sources). That is a large sample of documents rather than a controlled study of what happens in the room. Our belief, after that many, is that a report survives questioning when every input traces back to something named and dated. In a survey of 885 institutional VCs at 681 firms, Gompers, Gornall, Kaplan and Strebulaev found that 9% of all VCs and 17% of early-stage investors use no quantitative deal evaluation metric, and that 20% of all VCs and 31% of early-stage VCs do not forecast cash flows. So roughly a third of early-stage investors build no cash-flow model of their own, and the only one on the table may well be yours.
It is also the argument for bringing the assumption set before the term sheet rather than reacting to one. Galinsky and Mussweiler (2001) found across three experiments that whichever party made the first offer obtained a better outcome. The useful reading of that is not “name a big number first”. A good valuation is not the highest one you can get away with; it is the one that closes the information gap between the founder who knows the business and the investor who knows the market, and sets up a partnership on terms both sides understand. What you want in the room first is your reasoning, not your price.
How to build one, in order:
- Gather the financials: P&L, cash flow and balance sheet, with their growth and cost drivers written down. For a pre-revenue company the model usually covers about three years. Stretching further captures more growth but buys more uncertainty with it, and mainly earns its place when the product launch sits beyond year three.
- Run the five valuation methods. Two qualitative, three quantitative, each calculation shown, combined into a weighted average.
- Benchmark the result. Compare your implied multiples to a peer set matched by stage, region and industry.
- Appendix every parameter. Each input with its source, date and update frequency.
- Write the two-paragraph summary. The range, how the weighting was set, and the assumptions doing the most work.
What goes in a startup valuation report?
A startup valuation report needs six sections: key information, financial data, the valuation methods with their full calculations, supplemental deal information, benchmarking against a peer set, and an appendix listing every parameter with its source. For a template to work from, ours runs to 34 pages in exactly that order, with the benchmarking pages appended for the benchmarked version, and there is a complete sample report: a real example, filled in end to end.
| Section | What it contains | The question it answers for the VC |
|---|---|---|
| Key information | Company snapshot, stage, sector, geography, headline range | “Is this in my mandate?” |
| Financial data | Three-year P&L, cash flow and balance sheet forecasts, longer where launch is further out, with their drivers | “Do the numbers follow from the strategy?” |
| Valuation methods | One page per method: inputs, parameters, full calculation | “Can I reproduce this?” |
| Supplemental information | Cap table, previous rounds, equity holders, use of funds, round size | “What does my money buy, and who else is on the register?” |
| Benchmarking | Implied multiples and financial metrics against a relevant peer set | “Is this a sane range for companies like this one?” |
| Appendix | Every parameter, its source, update frequency, and any adjustment from default | “Where did each input come from?” |
The appendix tends to be the last section a founder writes and the first an investor turns to. A discount rate with no derivation is an opinion; one with a risk-free rate, a country market risk premium, a beta and a dated source behind it is something you can argue about.
How long should a startup valuation report be?
Long enough that every input is traceable. Ours runs to 34 pages, most of it not narrative but calculations, statements, benchmarking pages and the appendix.
Sequencing matters more than length. Send the two-paragraph summary with the report attached, and hold the granular model, cohort data and cap table for diligence. An investor who wants to check your discount rate will ask, and answering it then beats sending the whole data room up front.
Is a 409A valuation the same as a fundraising valuation?
No, and conflating them is the fastest way to look unprepared. A 409A prices common stock so you can grant options at a defensible strike price for tax purposes. A fundraising valuation prices the round: preferred shares carrying liquidation preferences, conversion rights and sometimes participation or dividends. The two are different by construction, though they should still be coherently linked and rest on the same underlying view of the business; a 409A that implies a different company than your fundraising model is a problem of its own. The gap also reflects a discount for lack of marketability, which a16z notes most studies put in the 25–35% range for a two-year holding period, while cautioning that “there is no ‘rule of thumb,’ and every company is unique.”
So do not hand a VC your 409A and call it a valuation.
Why five methods instead of one?
Because any single method carries the bias of its own assumptions; Equidam combines five (two qualitative, three quantitative) into a weighted average. A DCF is highly sensitive to its terminal value; a multiple is a bet that the market prices your peers well. Combining several gives you a range you can defend, and stops one bad input from deciding the answer.
The two qualitative methods are the Scorecard Method (team, opportunity, product, competition, channels and funding needs, scored against a comparable baseline) and the Checklist Method (value assigned against risk-reduction milestones achieved). They matter because of what VCs say they select on. In the same Gompers survey, the management team was named an important factor by 95% of VC firms and the single most important factor by 47%, ahead of business model (83%), product (74%) and market (68%).
The three quantitative methods are DCF with Long-Term Growth, DCF with Multiples, and the Venture Capital method, which divides an estimated exit value by the return the investor requires to get there. Everything hangs on that required return, and it is stage-specific: it falls as the company matures, so a seed hurdle and a Series B hurdle are not interchangeable. The method earns its place because it is the language investors evaluate in: in that survey, cash-on-cash multiples were the most popular metric at 63% of the sample and IRR second at 42%, while only 22% used NPV methods.
The same paper reports a mean required multiple of 5.5x and a median of 5x. Read that as a fund-level average across all the stages the survey covers, not as a pre-seed input. The authors say so themselves, noting “higher multiples for early-stage and small funds”, and the required return applied at the earliest stage is several times what a Series B investor works with. Divide your exit value by five at pre-seed and you are quietly using a later-stage investor’s hurdle rate.
The weighting is determined by stage. Qualitative methods carry more weight when there is no financial history, quantitative methods take over as performance data accumulates, and the qualitative methods drop out entirely at growth and maturity.
Three startup-specific adaptations keep the quantitative side honest, each belonging in the appendix with its source: survival rates applied year by year to projected cash flows (ours from Eurostat and the US Bureau of Labor Statistics), an illiquidity discount (ours drawn from our own research), and a beta adjusted for stage, size and profitability rather than industry alone. The pattern is the same for every one of them: named source, update frequency, no orphan numbers.
Comparables are the one method to keep out of the driver’s seat. Market multiples are pro-cyclical, one-dimensional and easy to game, and they handle genuinely new propositions badly (the reasoning, in full).
Which assumptions get attacked first?
Almost always the ones connecting revenue to cost. Before you send anything, check your projections for the six patterns that draw the first questions:
- Narrative and numbers disagree. The pricing in the model doesn’t fit the segment in the pitch.
- Missing growth-related costs. Revenue 5x, sales and marketing 2x, nobody hired to service it.
- Unexplained inflection. Growth jumps in year three with no milestone, channel or hire behind it.
- Implausible penetration. A large share of a large market on a small budget.
- Self-improving unit economics. Margins climbing with no pricing or cost change behind them.
- Reverse-engineered growth. Rates calibrated backwards from a valuation you picked first, usually visible in how round the numbers are.
Then add the two things that turn a forecast into an argument. Sensitivity analysis: vary three to five drivers (acquisition cost, conversion, average revenue per user, churn, gross margin) across a defensible range, and show which ones actually move the valuation. And three scenarios: base, upside and downside, we’d suggest roughly 30–50% either side of base, the upside built from correlated improvements rather than a wish list, the downside testing whether the business survives and answering “what if the main acquisition channel underperforms?” Build the model with our free financial projections template.
Where does benchmarking data come from, and what does it miss?
Benchmarking is where you take the implied multiples from your valuation (value over ARR, value over EBITDA) and check them against companies like yours. It checks the output of your valuation, not the input. The hard part is choosing the peer set.
Two sources dominate, and both are explicit about their scope. Carta publishes private-market data drawn from the companies on its cap-table platform, weighted towards US startups. The PitchBook-NVCA Venture Monitor describes itself as “the definitive review of the US venture capital ecosystem”, and its methodology draws the boundary in one sentence: “All financings are of companies headquartered in the US, with any reference to ‘ecosystem’ defined as the combined statistical area (CSA)” (Q2 2026 Venture Monitor, p. 33). Both do exactly what they say. The question is whether what they say covers you.
Both describe priced US venture rounds, and two things follow. First, they only contain priced rounds. At pre-seed almost all of the money moves on SAFEs and convertible notes — Carta’s own pre-seed data puts SAFEs at 93% of pre-seed rounds in Q1 2026 — which set a cap rather than a valuation, so there is nothing for a priced-round dataset to record. Whatever the earliest stage actually looks like, it is thin in this data by construction. That is a gap in the data, though, not permission to skip the work: a cap is still a number that has to be arrived at and defended. An intuited cap gets taken apart the same way an intuited pre-money does, and unlike a pre-money it sets the reference point your priced round is negotiated against, months before anyone runs the model. Everything in this article applies whether the instrument you sign carries a valuation or a cap.
Second, headline valuations are prices for preferred stock with protections attached. Modelling 135 US unicorns, Gornall and Strebulaev found reported post-money valuations averaged 48% above fair value, with common shares 56% overvalued. Adjusting for terms such as IPO return guarantees and down-IPO vetoes, 65 of the 135 lost unicorn status. Benchmark against a headline number and you import someone else’s deal terms into your own.
Raising outside the US widens the gap further. In H1 2026, megadeals of $100 million or more made up 87.5% of the $412.7 billion deployed, and AI companies took $355.9 billion of it, 86% of all venture dollars in the period (Q2 2026 Venture Monitor). A median drawn from that distribution reads a market dominated by very large late-stage AI rounds. A European, Middle Eastern or Latin American pre-seed round is priced against a different cost base, exit market and pool of capital, none of which shows up in it.
We can put a size on that difference, because we measure it. In our H1 2026 Valuation Delta, built on 3,000+ pre-seed valuations across seven regions, the median pre-seed valuation is $5.16M in the United States against $3.30M in Europe, $2.99M in Africa and $2.40M in Latin America. Implied dilution splits the same way: 26.9% in the US, 17.7% in Europe. So when a founder in Lisbon or Nairobi is shown a US benchmark and told the ask looks high, most of what they are being shown is the distance between two markets, not a problem with their number. Higher dilution is not weakness either; it tracks the size of the raise against the valuation, not the quality of the company.
The global figure makes the point from the other side. The all-region median pre-seed valuation is $5.20M, and it sits above six of our seven regional medians, dragged up by the US. Any single headline number does that. It is the wrong number to hand a founder who is not raising in the market that set it.
That regional split is the gap our dataset fills. The benchmark set behind those 160,000+ valuations is matched by stage, region and industry rather than drawn from US rounds several stages ahead of you, alongside 30,000+ public-market comparables refreshed weekly (Equidam data sources). The Valuation Delta hub carries thirteen editions of that data going back to Q2 2023, so the half-year you are looking at can be checked against the ones before it.
Use each source for what it covers, then bring a peer set that contains companies like yours, name it, and say why. Present it as context rather than constraint: show your growth, margin and multiple against the median and 25th/75th percentiles, and say plainly where you sit.
How do you defend a startup valuation to investors?
Four habits, most useful first.
Present a range, not a point. The weighted average is the point estimate, so the range has to come from somewhere else: the spread of the five method outputs before they are combined. Show what each method produced. The distance between the lowest and the highest is the honest width of the estimate, and it is the part a single counter-number cannot answer.
Take the argument to the assumptions. If a VC says the valuation is too high, the useful reply is “which assumption do you think is wrong?” Market growth, conversion, cost structure, exit multiple, required return: each is discussable, and each is already in your report. This is where the anchoring research earns its keep. Galinsky and Mussweiler found the first-offer advantage disappeared once the other side focused on information inconsistent with the anchor, which is what a VC does to a number with no derivation behind it.
When their comparable disagrees with your model, ask what it’s comparable to. Stage, geography, capital structure and date all matter. A US Series A priced 18 months ago is not a read on a European pre-seed today. Keep it a question about scope, not about the quality of their data.
Run the implied-multiple check last, out loud. Doing it in front of them shows you stress-tested your own number first.
Begin with the assumptions
A valuation is an estimate produced under stated assumptions. It is never right or wrong in the abstract. It is defensible or it isn’t, and what makes it defensible is that every input is visible, sourced and dated. That transparency also lets founders pitch the harder idea rather than the legible one. In Startup Catering to Venture Capitalists, Xiyue Li found that information frictions in valuation push startups to select projects matching the expertise of the investors they are pitching, and that these catering projects were 19.3% less likely to get patent approval.
If you want to build this yourself, the structure above is the whole specification. The test is simple: for any number in the document, a reader should be able to find where it came from and when it was last updated without asking you. If you would rather not assemble it by hand, that report is what Equidam produces, and it carries 94% positive investor feedback (Equidam internal data). Either way, the best startups are valued, not priced, and the difference shows up in what you can explain.