The Deal Screen Standard: What a Company Must Prove Before Partner Review
You can grade any sourced company against fixed thesis gates and verified-evidence checks, produce a pass-or-advance verdict, and have a second associate reproduce it.
This standard is for associates, platform and talent partners, and angels who decide whether a sourced company is verified and qualified enough to put in front of a partner. It gives you a fixed set of thesis-fit gates and public-evidence checks, so your pass-or-advance verdict holds up and a second associate grades the same case the same way. The goal is to stop passing gut calls up the chain, and to stop partners re-verifying every deal from scratch.
Existing guidance either lists soft traits VCs supposedly look for, or walks through the multi-week diligence funnel. Neither states a gradeable bar for the single gate between a sourced company and a partner's calendar. That gate is this document.
Why the first screen is the decision that matters
Venture funds decline roughly 99% of deals, and most of those nos happen before a formal investment committee decision. The first screen, not the IC vote, is where the bulk of a fund's rejection decisions are actually made. That makes its reproducibility the thing that governs pipeline quality.
The aggregate funnel is unforgiving. Out of the 1,000 to 2,000 deals a fund may source in a year, about 1% reach a term sheet. One documented funnel screens 500 to 800 decks for every 50 to 100 conversations, evaluates around 20 opportunities with serious intent, and funds just one or two a year. The screen is the widest, cheapest gate in that funnel, and it is the one an associate owns.
Because so many decisions happen here, the failure mode is not being too harsh. It is being unreproducible. If an associate's pass depends on how the founder's deck made them feel, a colleague reviewing the same record cannot reach the same call, and the partner has to re-verify everything. A gradeable standard fixes that. Two people should read one record and produce one verdict.
Where the screen sits in the sourcing funnel
- 500-800Decks screened
per year in one documented funnel
- 50-100Conversations
first meetings held
- 20Serious evaluation
opportunities with real intent
- 1-2Funded
term sheets that close
The thesis gate: five dimensions, graded before anything else
The first thing to check is thesis fit, and it is a hard gate. A published fund thesis specifies five dimensions: stage, sector, geography, check size, and the thematic belief that binds its deals together. Grade a sourced company against all five before spending a minute on traction.
Misfit is an automatic pass. A misaligned thesis equals an automatic no, regardless of traction or team quality, because funds have specific mandates from their limited partners and cannot invest outside them even when they love the company. Thresholds are stated as bands tied to fund size: a $500 million fund might target Series B and C, while a $50 million fund concentrates on seed and Series A. Name the specific stage. VC Lab guidance instructs managers to target one investment stage, such as pre-seed, seed, or Series A, and to avoid vague terms like "early stage".
Grade each dimension pass or fail with the threshold cited. "Sector: fintech, thesis targets fintech and vertical SaaS - pass" is gradeable. "Feels adjacent to our thesis" is not. If any of the five fails and the failure is structural rather than a judgement call, stop. A large share of opportunities can be eliminated here in minutes, which is the point.
Verifying claims against public sources
Verification is a short desk pass that tests whether the company's load-bearing claims survive a look at the public record. Investors spend roughly 30 to 45 minutes on desk research before deciding whether to take a first call. That is cheap enough that skipping it is a procedural failure, not a resource one.
Each claim type maps to a specific source. Confirm the round and cap table against public funding and registration records. Confirm the founding team's history and headcount against public LinkedIn records. Confirm corporate registration against SEC EDGAR in the US or Companies House in the UK. Test technical claims against published research. The discipline is to mark each load-bearing claim confirmed, unconfirmed, or contradicted, with the source you checked noted beside it.
| Claim to verify | Where to check it | What the check proves |
|---|---|---|
| Funding round and history | Public funding and deal records | The round exists and matches the pitch |
| Team background, headcount | Public LinkedIn records | Founders have the history claimed |
| Corporate registration | SEC EDGAR (US), Companies House (UK) | The entity is real and filed |
| Web traffic claims | Public traffic estimators | Usage claims are plausible in scale |
| Technical or research claims | Published research indexes | The science is real, not marketing |
The trap is that most databases relying on self-reported or scraped financials cannot confirm the one claim that usually matters. A revenue field in a database is not a source-of-record confirmation. Read it as unconfirmed unless it is backed by a filing, by payment-processor data, or by a number you pulled yourself. Note the source type next to every claim so a reviewer can see the difference between "confirmed by a filing" and "the founder told a database."
Referral quality is graded evidence, not metadata
Log the intake channel and referral quality, because channel is predictive evidence in its own right. Warm intros convert to a first meeting at 20 to 30% and to a term sheet at about 1 in 10. Cold email converts at 1 to 2% and about 1 in 50. That is a roughly tenfold to thirtyfold gap in first-meeting conversion for the same funnel stage.
| Channel | To first meeting | To term sheet |
|---|---|---|
| Warm intro | 20-30% | ~1 in 10 |
| Cold email | 1-2% | ~1 in 50 |
| Aggregate (all sourced) | not published | ~1% |
A warm intro from a respected operator is not proof the company is good, but it is a real prior that belongs in the record as a graded field. Tag every deal inbound, outbound, or warm intro, and name the referrer when warm. When you write the one-pager, the channel line should sit near the top, because a partner reading it wants that prior before the metrics.
Finding the right people to source these deals, and the operators who make warm intros credible, is its own search problem. When you need to identify partners, founders, or domain experts across public professional records, describing them in plain language beats stitching together filters by hand.
The screening procedure, step by step
Run the screen in a fixed order so the record is complete and reproducible. The thesis gate comes first because it is the cheapest disqualifier; verification and plausibility follow; the memo and verdict come last. Every step has a definition of done so a reviewer can tell whether it was actually performed.
The seven-step first screen
- Run the thesis-gate check firstGrade stage, sector, geography, check size, and business model against the fund's published thesis before any other work. Done when every gate is marked pass or fail with the threshold cited.
- Log the source and referral qualityRecord the intake channel and referral quality, since channel predicts conversion. Done when the source is tagged inbound, outbound, or warm intro, with the referrer named if warm.
- Run the desk verification passConfirm round, traction, and team against named public sources in a 30 to 45 minute pass. Done when each load-bearing claim is marked confirmed, unconfirmed, or contradicted with the source noted.
- Run the plausibility and consistency testCross-check headline metrics against headcount, web traffic, and visible customers. Done when there is no unexplained inconsistency, or the inconsistency is flagged with the numbers that clash.
- Scan for the standard decline reasonsScreen for unsourceable TAM, traction inconsistent with observable signals, and a team with no verifiable domain connection. Done when each of the three categories is checked and cleared or triggered.
- Write the screening one-pagerProduce a one to three page brief stating what the company is, why it fits, what must be believed, and what to validate. Done when it answers whether the deal is worth a partner meeting, not whether to invest.
- Record the verdict and route itProduce pass-or-advance with a named champion if advancing. Done when the verdict is recorded with its reason category and, where the firm runs one, has passed a shadow IC.
The order is load-bearing. If you write the memo before the verification pass, you write persuasion instead of a screen. Some firms insert a shadow IC before the verdict, where junior team members stress-test the deal before formal presentation. Where that exists, it is step seven's gate, not a separate stage.
How the screen goes wrong: false positives to catch
The most valuable part of a screen is knowing where it lies to you. Each of the failures below is a case where a claim looks like it passes but does not. For every criterion you check, know what it proves and what it looks like when it deceives.
A claim that survives a headline read and dies on a definition read is the one that costs you a partner meeting.
Thesis gate passed on marketing copy, not behavior
A fund's website says "seed" but its last 18 months of activity are all Series A. The thesis you graded against was the marketing thesis, not the deployment reality. Portfolio data older than 18 months often points to a fund in harvest mode, not active deployment. Check recent activity, not the pitch page.
Traction confirmed on a headline number
"$50k MRR" often means the company did $50k total last month and $40k of it was one-time setup fees. A headline confirmed is not a metric verified. Check the revenue definition and its recurrence before you mark it confirmed.
TAM accepted from a press release
An impressive market number cites a press release or no source at all. When you search for the underlying data, the number cannot be confirmed. Search for the primary data; if it cannot be found, mark the TAM unconfirmed and say so in the memo.
User count without active users
"10,000 users" might be technically true, but ask about actives in the last 30 days and it becomes 147. Registered is not active. Ask for the 30-day active count and grade on that.
Signed contracts that are pilots
"Three signed enterprise contracts" turn out to be three pilot agreements with tiny budgets and no commitment to renew. A signature is not a commitment. Check contract value and renewal terms before treating logos as revenue.
The decision model sets the evidence bar
Match the screen's evidence bar to how the firm actually decides. The decision model determines how much a passing screen must prove, not the other way around. Under the champion rule, the screen needs to arm one advocate. Under unanimous consent, it must pre-empt every partner's likely objection.
A majority of surveyed firms use the champion, or sponsor, voting rule, where the committee proceeds as long as at least one partner champions the deal. This rule is used primarily at seed and early stage. Later-stage firms move toward consensus-based voting. Sequoia and Benchmark are documented unanimous-consent outliers, where every partner, regardless of seniority, can veto.
| Model | Who can advance a deal | Documented example |
|---|---|---|
| Champion / sponsor | One conviction partner | Majority of early-stage firms |
| Majority vote | Enough voting members | Common at growth and late stage |
| Unanimous consent | Every partner, any can veto | Sequoia, Benchmark |
| Partner checkbook | Single GP, up to a check-size cap | Polled firms |
The practical consequence: the same company screened for a champion-rule fund can advance on evidence that would not clear a unanimous-consent fund. If you work at a Sequoia-style firm, your memo must anticipate the objection of the most skeptical partner, because that partner holds a veto. Write the bar into your standard so the whole team calibrates the same way.
Calibrating the screen to the decision model
The screening one-pager
Write a screening memo of one to three pages before the company meets the full partnership. Its job is to decide whether the deal is worth a partner meeting, not whether to invest. Keep it short on purpose. The published one-page, six-section investor memo format exists because the constraint forces you to state only what is load-bearing.
The memo must answer four questions a partner will ask in the first thirty seconds: what the company is, why it fits the thesis, what must be believed for this to work, and what to validate next. Attach the graded verification table so the partner sees which claims are confirmed and which are not. That table is what stops the partner re-verifying from scratch.
COMPANY: name, one-line description, stage, geography SOURCE: channel (inbound / outbound / warm intro), referrer if warm THESIS FIT: stage [pass/fail] | sector [pass/fail] | geography [pass/fail] | check size [pass/fail] | thematic [pass/fail] WHAT IT IS: two sentences, plain English WHY IT FITS: the thematic belief this deal expresses VERIFICATION TABLE: - claim :: source checked :: confirmed / unconfirmed / contradicted - claim :: source checked :: confirmed / unconfirmed / contradicted - claim :: source checked :: confirmed / unconfirmed / contradicted WHAT MUST BE BELIEVED: the one or two leaps the case rests on TO VALIDATE NEXT: what a first meeting should confirm VERDICT: PASS or ADVANCE | reason category | named champion if advancing
Keep to one to three pages. Fill only what is load-bearing; delete rows that do not carry weight.
Record the verdict with its reason category. Where a taxonomy is useful, the three reusable first-pass decline categories are: TAM that cannot be sourced, traction inconsistent with observable signals, and a founding team with no verifiable connection to the problem domain. Bessemer's public Anti-Portfolio, a catalog of declined deals and the reasoning behind each pass, is a model for treating pass reasons as reusable knowledge rather than one-off notes.
Verify before you call it done
Run this checklist before you route a verdict. Each item is a checkable statement, not a topic, so a reviewer can confirm it was done. If any item cannot be checked, the screen is not finished.
Screen completion checklist
- All five thesis dimensions are graded pass or fail with the threshold cited.
- The intake channel is tagged and the referrer is named if the deal came warm.
- Every load-bearing claim is marked confirmed, unconfirmed, or contradicted with the source noted.
- The source type (filing, processor-backed, self-reported, scraped) is recorded beside each metric.
- Headline metrics have been cross-checked against headcount, web traffic, and visible customers.
- The three standard decline reasons have each been explicitly checked.
- Any revenue claim has been tested for recurrence, and any user count for 30-day actives.
- The one-pager states what the company is, why it fits, what must be believed, and what to validate.
- The verdict is recorded with its reason category and a named champion if advancing.
The reproducibility test is the final gate: hand the record to a second associate and see whether they reach your verdict. Depth of team makes this easier or harder. In Refolk's index of professional profiles, there are 1,073 US professionals in venture investment roles against 164 in the UK, a US pool roughly 6.5 times the UK's. A US associate can staff a shadow IC or a second reviewer more easily; in thinner markets the second-associate reproducibility standard is structurally harder to meet, which makes a written, gradeable standard more important, not less.
Keeping the standard current
Adopt this as team policy, then re-check the two inputs that drift. First, re-confirm the fund's live thesis against recent deployment, not the website, since portfolio data older than 18 months can signal harvest mode rather than active investing. Second, re-check the decision model whenever the partnership changes, because a shift from champion to consensus voting raises the evidence bar for every deal you screen.
Treat pass reasons as a growing catalog. Log each decline with its reason category so patterns surface over time, the way a published anti-portfolio does. When you find a new false positive that fooled the screen, add it to the failure-modes section and to the checklist. A standard that captures its own misses gets more reproducible with every deal it grades, and that reproducibility is the whole point of putting the gate in writing.
Questions practitioners ask
What is the difference between a screening memo and an IC memo?
A screening memo, also called a deal brief, is short at one to three pages and is written before the company has met the full partnership. Its job is to decide whether the deal is worth a partner meeting, not whether to invest. The IC memo comes later, after diligence, and argues the investment decision itself. Do not conflate them: the screen answers advance-or-pass, the IC memo answers fund-or-decline.
When should I take a first meeting with a startup?
Take the meeting once the company passes every thesis gate and its load-bearing claims survive a 30 to 45 minute desk verification pass without an unexplained inconsistency. Thesis misfit is an automatic pass regardless of how strong the team or traction looks. If a headline metric contradicts observable signals like headcount or web traffic and cannot be explained, decline rather than spending a partner's calendar on it.
How much of a fund's deal flow gets rejected at the first screen?
Venture funds decline roughly 99% of deals, and most of those nos happen before a formal investment committee decision. That means the first screen, not the IC vote, is where the majority of rejection decisions are actually made. Because so much of the fund's judgment lives at this gate, the reproducibility of your screen governs pipeline quality more than the committee vote does.
Why does referral quality belong in the screening record?
Because it is predictive evidence, not metadata. Warm intros convert to a first meeting at 20 to 30% and to a term sheet at about 1 in 10, while cold email converts at 1 to 2% and about 1 in 50. That is a roughly tenfold to thirtyfold difference in first-meeting conversion, so the channel a deal arrived through carries real signal and should be a graded field in the one-pager.
Can I treat a revenue figure from a startup database as verified?
No. Most databases rely on self-reported or scraped financials, so a figure appearing in one is not a source-of-record confirmation. Note the source type beside every claim and reserve confirmed for filings, payment-processor-backed data, or an active-user pull you ran yourself. A claimed MRR should be tested against its recurrence and against headcount and traffic before it counts as verified.