Investment Thesis Template for VC Funds
Three documents and two sheets that turn a thesis into one number: the share of your own market map you have to meet.
Free download · No account needed
Thesis Statement
[Fund] — [Sector] Investment Thesis
Each claim is written so that it could be refuted, carries the evidence accumulated on both sides, and names the observation that would settle it.
Claims
| ID | Claim | What would refute it | State |
|---|---|---|---|
| C1 | — | — | — |
| C2 | — | — | — |
| C3 | — | — | — |
| C4 | — | — | — |
| C5 | — | — | — |
| C6 | — | — | — |
Claim state has four values: supported, contested, refuted, and untested. Untested after four quarters of sourcing is a finding about where the fund has been looking rather than a gap in the research.
Rewrite conditions
The conditions under which this document gets rewritten rather than edited, written down now while nobody is defending it. A thesis with no rewrite condition survives its own refutation.
Stated assumptions, not tracked here
The beliefs the fund holds that this document is not testing, listed so a later reader does not mistake silence for evidence.
Scope
What the thesis does not claim, so nobody extends it into a market it never covered.
A 60 million dollar seed fund needs 24 investments. Its blended conversion rate of 1.03 percent would mean seeing 2,327 companies, against a thesis universe of 457, so the thesis looks 5.1 times too narrow. Re-scored on the in-scope cohort of its own prior fund the rate is 5.93 percent, and the requirement is 405 companies. The thesis is investable, but only at 88.6 percent coverage of its own market map. That number appears in no thesis document anywhere.
Page one for this query is unanimous. Every result hands you the same fill-in-the-blank sentence: fund name, size, stage, check range, geography, sector, secret sauce. The better ones add a line for invalidation, which is a blank rather than a mechanism, because nothing downstream ever reads it. None of them counts the universe the sentence defines, and none divides that count into what the fund model needs. The same gap runs through screening inbound deals, where the criteria get stated and the thresholds do not.
Then one claim is refuted, at one of twelve against a floor of three of twelve. 96 companies leave the universe, required coverage goes to 112.2 percent, and the thesis supports 21 investments rather than 24. Committed capital has not moved, so the average check goes from 1.45 to 1.66 million dollars and ownership from 12.1 to 13.8 percent. A claim an adviser must be able to substantiate on demand belongs in the methods of analysis it already discloses.
What is in the pack
Thesis Statement
Scope separated from claims, then four to eight claims each carrying a number, a measurement window, the observation that refutes it, and what it makes addressable in companies or dollars.
Market Participant Register
Every company in the category, then marked in or out of scope with the reason. Carries founded year, so the formation rate is computable, and wedge, so a refuted claim re-scopes the universe by filter rather than by project.
Thesis Evidence Tracker
One row per dated observation, never one per conclusion, with the measured value in the threshold's own unit and a column marking which observations can actually resolve it.
Market Analysis
Where every figure is computed. The counted universe, both conversion rates with the arithmetic shown to close, required coverage, coverage by segment, claim states, and the chain each refutation sets off.
Sourcing Implications
The monthly in-scope first-contact rate, the segment plan ordered by companies unmet, which channels reach which segments, and the screening standard each claim implies for a first call.
A claim without a threshold is a preference
The space rule. Three claim states rather than two, with at-threshold named separately because that is where a decision is still cheap. The same one-way reading of evidence governs a pipeline stage nobody has touched.
How it works
- 1
Separate scope from claims
Scope is who you will look at, and it needs no threshold. Claims are what you believe about that scope, and each one gets a number, a window and the observation that refutes it.
- 2
Count the universe, do not sample it
Add every company in the category first, then mark each in or out of scope with the reason. Read the formation rate off the founded-year column, then reconcile against standing count plus formation.
- 3
Re-score your own funnel
Your blended rate was earned under a broader mandate. Re-score the prior fund's record against the new scope to get the in-scope rate, which in the worked example is 5.93 percent against a quoted 1.03.
- 4
Divide, then read the band
Required investments over the in-scope rate, over the available universe. Under 40 percent means selection is the constraint, 40 to 90 means sourcing is, and above 100 means the fund is too large for it.
Frequently asked questions
Is a refutation threshold not just an invalidation line, which the good templates already have?
The line exists on some of them. What is missing is anything reading it. Here a refuted claim re-scopes the register, which changes the available universe, the supportable investment count, the average check and the ownership target. In the worked example that chain runs from one diligence finding to 1.7 points of ownership.
Where does the in-scope conversion rate come from if this is a first fund?
From angel or operator deal records, or from a stated assumption flagged as one. The reconciliation still works, it just returns a range instead of a number. Running it on the blended rate instead makes any narrow thesis look impossible, so read that version as a floor rather than an answer.
Why reconcile against a universe that includes companies which do not exist yet?
Because a four-year investment period sources from them. In the worked example the standing count is 269 and formation adds 188. Reconciling against 269 alone would report 150.6 percent required coverage and kill a thesis that is investable. Formation is read off the founded-year column, across complete years only.
What if required coverage comes back above 100 percent?
Then the thesis cannot support the fund at that size, and there are three levers. Widen the scope, which lowers the in-scope rate and has to be re-reconciled rather than assumed to help. Win more of what you already see. Or make fewer, larger investments, which is what the arithmetic already says.
Does this rank companies or tell me which ones to invest in?
No. It counts a universe, reconciles it against your fund model, and reports what has accumulated against each claim. The next-action column says source, monitor, interview or ignore on scope grounds only. Judging a specific company is what the memo is for, and the partnership decides.
How is thesis drift different from just reviewing the portfolio?
It scores deployed capital against the claims as written, one at a time, and reports the share by capital as well as by count. Those diverge whenever the off-thesis deals were larger. In the worked example 60.0 percent by count is 60.6 percent by capital. Ongoing reporting then tests the claims themselves.
Can I use the market map without writing the claims first?
You can build it, but it will not be a denominator. Scope is what decides which rows are in scope, so without written claims the register is a list of interesting companies rather than a count. That is the failure mode a stated screening criterion also exists to prevent.
Find out what coverage your thesis is quietly demanding
Send the draft thesis, whatever company lists you have, and your screening record with outcomes. The first pass counts the universe and divides your fund model into it.
Reconcile my thesis