River
Y CombinatorBacked by Y Combinator
FREE TEMPLATE

Deal Screening Criteria Template

Three documents and three sheets that screen on three values, so a figure the deck never stated stops counting as a figure that missed.

Free download  ·  No account needed

Screening Criteria

[Fund] — Inbound Screening Criteria

Every criterion resolves to yes, no, or the deck did not say. The third value is the one this space exists to keep.

Stage 1. Mandate gates

A deck failing any gate is declined on stated facts, the same day, with no follow-up question attached.

GateThresholdDefinition
Stage and round sizeThe round being raised now, not the total ever raised
SectorThe thesis as written, not the partner’s current interest
GeographyWhere the team and the customers are, not where the entity is registered

Stage 2. Threshold criteria

CriterionThresholdDefinition that makes it checkable
Recurring revenue
Growth
Net dollar retention
Gross margin
Burn multiple
Founder evidence

Write the number even where it feels arbitrary. One arbitrary threshold applied to sixty decks a week is a policy. A vague one applied sixty times is sixty policies.

The advance rule

How many criteria must be met, and how many may be unmet, before a deck advances. Fixed before the week starts rather than argued deck by deck.

Sixty decks arrive in a week. Twenty-seven decline at the mandate gate on stated facts. Thirty reach the threshold criteria, where 72 of the 180 criterion cells were never stated at all. Two advance, thirteen decline, and fifteen cannot be decided either way. On a scorecard those fifteen score low and get binned with the three unclear gates, which makes eighteen of the week's fifty-eight declines an absence rather than a fact.

Every screening scorecard rates each criterion one to ten, weights the dimensions and totals them. A company with a 41 per cent gross margin scores low. A company whose deck never mentioned gross margin scores low too, because there is nothing else to type. The same collapse shows up in portfolio reporting, where an absent figure is not a bad quarter. This screen has a third value and a rule that uses it: a verdict is a verdict only when no assignment of the missing fields would change it. Everything else is a request.

That same rule also cuts the number of requests. Twenty-eight of the thirty screened decks carry a blank; only fifteen need an email, because in thirteen the answer could not have moved the verdict. The mandate gates are not preference either. Rule 203(l)-1 caps a venture capital fund at 20 per cent of its committed capital in assets that are not qualifying investments, and the Marketing Rule expects an adviser to substantiate a material claim on demand.

One week of inbound, and the fifteen decks a scorecard would have declined without deciding

The Screen Result, the Missing Information Log, the Pipeline Register and the notes that go back out.

Screen Result

Illustrative, for a fictional generalist fund screening the week of 20 April 2026. Thirty of sixty decks got past the mandate gate. Advance needs four or more met and at most one not met. Nothing here is scored.

DealCompanyRevenueGrowthRetentionMarginBurn mult.FounderVerdictWhy
D034Jarrowmont Softwaremetmetnot metmetmetmetAdvanceEvery criterion stated, five met
D041Quarrenden SoftwaremetmetmetmetabsentmetAdvanceAdvances even if the blank comes back failed
D004Dunsford Gridmetnot metabsentmetabsentnot metDeclineTwo not met. Both blanks are irrelevant
D030Ollerton Networksmetnot metabsentabsentnot metmetDeclineDeclines however the blanks land
D001Aldergate SystemsmetmetmetabsentabsentmetUndeterminedBoth met gives 6 met and advances. Both failed gives 2 not met and declines
D022Vexley Biometmetabsentmetnot metmetUndeterminedOne field decides it either way
D027Larchmont HealthabsentabsentmetabsentabsentmetUndeterminedFour blanks, two met, nothing failed
D011Marlbury Logisticsmetmetmetmetabsentnot metUndeterminedThe one blank is the whole decision
WEEK30 decks past the gate6 absent9 absent19 absent13 absent22 absent3 absent2 / 13 / 1572 of 180 cells never stated, 40 per cent

D041 and D004 are the two halves of the rule. D041 has a blank and still advances, because five criteria are met and one failure would not break the threshold, so there is nothing to ask for. D004 has two blanks and still declines, because two criteria already failed and no answer can undo that. Neither gets an email. Only the four undetermined rows do.

Missing Information Log

Every blank across the thirty screened decks, with the load-bearing test computed rather than judged. Hold every other cell fixed, let this field take each value, and see whether the verdict moves.

DealCompanyField not statedWhat to ask forChanges the verdictIn this email
D001Aldergate SystemsGross marginRevenue less all cost of delivery, support and implementation payroll inside itYesYes
D001Aldergate SystemsBurn multipleNet new recurring revenue and net cash burn, last two quartersYesYes
D004Dunsford GridNet dollar retentionCohort revenue then and now, churn includedNoNo
D004Dunsford GridBurn multipleNet new recurring revenue and net cash burn, last two quartersNoNo
D011Marlbury LogisticsBurn multipleNet new recurring revenue and net cash burn, last two quartersYesYes
D027Larchmont HealthRecurring revenueContracted subscriptions with twelve months or more remainingYesYes
D027Larchmont HealthGrowthRecurring revenue at two dates, same definitionYesYes
D027Larchmont HealthGross marginRevenue less all cost of delivery, last full quarterYesYes
D027Larchmont HealthBurn multipleNet new recurring revenue and net cash burn, last two quartersYesYes
D041Quarrenden SoftwareBurn multipleNet new recurring revenue and net cash burn, last two quartersNoNo
WEEK30 decks past the gate72 blanksAsks are inputs, never metrics43 load-bearing15 emails, median 3 fields

Twenty-eight decks carry a blank and fifteen get an email. A completeness view sends twenty-eight, and thirteen of those go to founders about to be declined for something the answer could not have fixed. The asks are written as inputs rather than metrics on purpose: a founder asked for their retention figure replies on their own definition, which is where the criterion already was.

Pipeline Register

One row per deck received. Five stages, five verdict phrasings, used consistently, because a register whose reasons are freely written cannot be counted.

DealCompanySourceReceivedStageOwnerVerdict basisNext action
D001Aldergate SystemsWarm intro20 AprAwaiting informationPartnerAbsent fields could produce either verdictRequest sent, all missing fields in one email
D002Barrowfield HealthCold inbound21 AprDeclined at gateAssociate 1Round size outside the 3m to 15m bandPass note sent
D004Dunsford GridCold inbound22 AprDeclined on the screenAssociate 2Declines however the absent fields come backPass note sent
D034Jarrowmont SoftwareCo-investor referral21 AprFirst meetingPrincipalEvery criterion stated, four or more metMeeting booked
D041Quarrenden SoftwarePortfolio founder referral23 AprFirst meetingPartnerAdvances however the absent fields come backMeeting booked
D047Westmere RoboticsAccelerator demo day20 AprDeclined on the screenAssociate 1Declines however the absent fields come backPass note sent
D055Ellerby AnalyticsCold inbound24 AprAwaiting informationAssociate 2Absent fields could produce either verdictRequest sent, all missing fields in one email
WEEK60 decksWeek of 20 April27 / 3 / 2 / 13 / 1527 declined at gate, 2 to first meeting, 13 declined on the screen, 18 never decided15 emails, one each

The source column is captured at arrival because nobody can reconstruct it later, and it is where sourcing quality eventually gets measured. Both first meetings this week came from referrals; neither came from the twenty-seven cold decks that failed the mandate gate.

Pass Note

Declined with blanks that did not matter. Sent to Dunsford Grid the same day, one of thirteen this week.

Thanks for sending this through, and for the detail in the deck. It made the read straightforward, which is not always the case.

We screen at this stage against six thresholds, and Dunsford came in below two of them on the figures in the deck: growth at 64 per cent against the 100 we look for at entry, and no prior operating role in freight among the founding team. Neither is a judgement about where the business gets to. Both are where our own model needs things to be when we come in.

There were two metrics the deck did not cover, retention and burn multiple. I checked whether either would have changed the answer before writing, and neither would have, so I have not asked you to go and pull them.

Good luck with the raise, and do come back if the growth picture changes materially. Send the quarter that shows it.

The bolded paragraph is the whole point of the category. It costs one sentence, it is only true because the log computed it, and it is the sentence founders remember three funds later.

What is in the pack

01

Screening Criteria

Two stages of criterion, each with the threshold as a number and the definition that makes it checkable, plus an advance rule a sheet can evaluate

02

Screen Result

One row per deck and one column per criterion, three values and no score, with the arithmetic behind each verdict written out in the row

03

Missing Information Log

Every blank with the load-bearing test computed, so a field the deck omitted only becomes an email when the answer would move the verdict

04

Pipeline Register

The object of record, with five stages, five verdict phrasings and source attribution captured at arrival rather than reconstructed later

05

Follow-up Question Set

One question per criterion, asking for the input rather than the metric, so a reply resolves the criterion instead of restarting the definition argument

06

Pass Note Templates

Four notes, one per way a deck leaves the funnel, each naming the actual reason instead of saying the timing is not right when timing is not the reason

07

The blank rule

A standing instruction that nothing in the space averages, weights or totals the three values, because one number cannot hold a fact and a gap at once

How it works

  1. 1

    Put numbers on the criteria

    Every criterion gets a threshold as a number and the definition that makes it checkable. Strong growth cannot be met or not met, so it becomes 100 per cent year on year.

  2. 2

    Screen on three values

    Met, not met, or never stated. A figure computed on a definition you cannot reconcile is never stated, not met. Nothing gets a score.

  3. 3

    Compute the verdict

    Enumerate every way the blanks could come back. If they all agree, that is the verdict. If they disagree, the deck is undetermined and the output is a request.

  4. 4

    Send one email each

    Only for blanks that move the verdict, with every one of them in the same message. Everything else gets a pass note or a meeting the same day.

Frequently asked questions

Why does a third outcome matter this much?

Because it is usually the biggest pile. In the worked week it is fifteen decks against two advances and thirteen declines. A scorecard cannot represent it, so all fifteen score low and leave the funnel looking like decisions when they were never decided. That undecided pile is what later fills a deal pipeline nobody trusts.

Will this generate more follow-up emails, not fewer?

Fewer. Twenty-eight of thirty screened decks carry a blank, and a chaser working from completeness emails all twenty-eight. The load-bearing test cuts that to fifteen, because in thirteen the missing figure could not have changed a verdict that was already settled.

Our criteria are not written down. Does this still work?

That is the normal starting point, and the first pass turns whatever you say into numbers and reads them back for correction. Correcting a proposed threshold is much faster than supplying one, and a criterion without a number can only be scored by impression. Where those numbers should come from is the thesis itself.

Is sixty decks a week realistic to screen properly?

Most of it costs almost nothing. Twenty-seven of the sixty fail a mandate gate on something the deck itself says, which needs no follow-up and no thought. Clearing those first is what leaves time to read the thirty that reached the criteria.

Does it decide anything or rank the deals?

Neither, and the boundary is deliberate. It sorts inbound against thresholds you set and names what to ask for. Nothing here scores or ranks companies against each other. Reading one deck closely is a separate job, which is what pitch deck analysis does.

What happens to a deck once it advances?

It leaves this space. The screen answers whether to take a first meeting, not whether to invest. What to actually ask in that meeting is a separate question, ranked by what the deck cannot answer, in the first-meeting question set. The document a partnership eventually votes from is a different artifact again, built by the investment committee memo pack.

What format are the downloaded files?

Word documents for the criteria, question set and pass notes, and CSV for the three sheets. The documents open in Word, Pages and Google Docs. The sheets open in Excel, Numbers and Sheets with the columns intact, and nothing needs converting.

Find out how many of last week's passes were never decisions

Send a week of inbound and your criteria, however rough. The first pass returns the three verdict counts and the decks that need one email each.

Screen my inbound