How to Build a VC Screening Rubric Partners Will Use
A VC screening rubric works when it asks a handful of separate questions your partners already care about, scores each one against written anchors before anyone gives an overall view, and records the reason behind every pass. It fails when it becomes a weighted score that people fill in after they have already made up their minds.
This post covers why most screening scorecards end up ignored, what research on structured judgment says to do instead, and a fill-in template with anchored levels, make-or-break checks and a reasons log.
Why most screening scorecards get ignored
A scorecard gets ignored when it records a decision rather than informing one. The usual pattern: a partner reads the deck, forms a view in a few minutes, and then the associate fills in ten scores that happen to agree with it. The numbers add work and no information.
Three design faults cause most of this:
- Too many criteria. Fifteen boxes invite guessing on the ten nobody has evidence for yet.
- No anchors. "Team: 4 out of 5" means something different to every partner, so scores cannot be compared across deals or people.
- One overall score. Adding weighted criteria into a single number hides why a company was advanced or passed, which is the information the fund most needs later.
What research on structured judgment says
The most useful guidance comes from Daniel Kahneman, Dan Lovallo and Olivier Sibony's "A Structured Approach to Strategic Decisions" (MIT Sloan Management Review, 2019). They describe a method they call the Mediating Assessments Protocol, built on three rules:
- Decide the assessments in advance. Identify "a handful" of attributes that are critical to the evaluation.
- Score each one independently and on evidence. People rating one dimension should not be influenced by each other or by other dimensions of the option.
- Make the overall judgment last. Discuss the whole option only after every assessment has been scored and shown together.
The roots are in hiring. When Kahneman designed an interview process for the Israeli Army in 1956, he replaced overall impressions with separate scores on six attributes. A simple average of those scores predicted performance better than an intuitive evaluation, and an intuitive prediction made after the separate scores were assigned did better still when combined with them.
The article also includes a venture example. A firm the authors call "VC Co." broke its criteria into separate assessments, with the founding team split into sub-assessments such as problem-solving skills, technical expertise and resilience. Each was rated on a percentile scale against a slate of founders the firm had met, and summary narratives arguing for an investment were banned in meetings until every assessment had been reviewed. The authors report the firm's early experience was that the process did not slow decisions, and they note that quantitative effects will take years to assess. Treat it as a documented practice, not a measured result.
Two details in the article go against common scorecard advice. The authors advise against assigning explicit weights, because it mechanizes the decision and leaves no room for informed intuition. And they say some dimensions should be framed as make-or-break assessments and evaluated first.
Put the team section first, and make it specific
Team criteria deserve the most care, because that is where early-stage investors actually look. In a randomized field experiment on AngelList, Shai Bernstein, Arthur Korteweg and Kevin Laws sent about 4,500 active early-stage investors nearly 17,000 emails that randomly showed or withheld information about a startup's team, traction and existing investors (working paper; published in the Journal of Finance in 2017). The average investor responded strongly to information about the founding team, but not to traction or lead investors. The least experienced investors responded to every category.
Two lessons follow for a rubric. First, "team" is too broad to score as one box; split it into the specific things your partners check, the way VC Co. did. Second, a rubric is a way to make an experienced partner's focus explicit, so newer team members do not give every piece of information equal weight.
The screening rubric template
Copy this structure and replace the anchors with your own. Everything below is an illustrative starting point, not a recommended standard; your criteria should come from the questions your partners actually ask.
Part 1: Make-or-break checks (answer before anything else)
| Check | Pass condition (write yours) | Result |
|---|---|---|
| Mandate fit | Stage, sector, geography and cheque size are inside the fund's mandate | Pass / Fail |
| Ownership and timing | The round size and timing leave room for the fund's target position | Pass / Fail / Unknown |
| Known deal-breaker | No fact that ends the review (for example, a conflict with an existing portfolio company) | Pass / Fail |
If any check fails, stop and log the reason. Do not score the rest.
Part 2: Independent assessments (score each separately)
Keep this to four to six assessments. Each gets a score, a one-line piece of evidence, and its source.
| Assessment | 1: Below our bar | 3: Typical of companies we meet | 5: Among the best we have seen | Evidence and source |
|---|---|---|---|---|
| Founder–problem fit | No direct experience with the problem | Worked near the problem or the customer | Has lived the problem and built in it before | |
| Technical or execution depth | Key work would need to be hired out | Team can build the first version | Team has shipped this kind of product before | |
| Market timing | No clear reason this works now | A plausible change makes it possible | A specific, checkable change has just opened it | |
| Early evidence of demand | None beyond the founders' view | Interviews, waitlists or pilots | Paying or repeat users, verified | |
| Thesis fit | Outside the fund's stated theses | Adjacent to a thesis | Directly on a current thesis |
The middle anchor is deliberate. Following the percentile idea above, "3" means typical of the companies your fund actually sees, not average for all startups. If you can, name past companies at each level so the anchors stay concrete.
Part 3: Overall view and reasons log (fill in last)
| Field | What to write |
|---|---|
| Overall view | Advance / Watch / Pass |
| Main reason | One sentence, tied to a scored assessment or a make-or-break check |
| What would change our view | The specific evidence that would reopen it |
| Open questions | Claims not yet confirmed independently |
| Reviewer and date | Who scored it, and when |
There is no total. The partners read the profile of scores and the evidence, then make the call. This follows the article's advice against explicit weights and keeps the reason for each decision visible.
How to run it in a weekly screening meeting
- Score before the meeting. Each reviewer fills in Parts 1 and 2 alone, so the first view spoken in the room does not set everyone else's scores.
- Review assessment by assessment. Show all reviewers' scores for one assessment, discuss gaps of two points or more, then move to the next.
- Hold the pitch until the end. Nobody makes the overall case for or against the company until every assessment has been covered.
- Decide and write the reason. Record the overall view and the main reason in Part 3 before moving on.
- Revisit passes each quarter. Look at which passed companies later raised and whether the logged reasons still look right. Change anchors when they keep failing, not after one miss.
Common mistakes
- Scoring what you cannot see yet. At first look, many fields will be unknown. Mark them unknown instead of guessing a 3.
- Letting the rubric grow. Each new box needs a reason. If a criterion has never changed a decision, remove it.
- Copying another fund's rubric. Your criteria should reflect your mandate and what your partners have learned. VC Co.'s list, the authors note, would not necessarily be the same for another venture capitalist.
- Hiding the reasons. A score without a logged reason is hard to learn from. The same point applies earlier in the funnel, when finding startups before they fundraise: keep the reason each company surfaced visible.
Where software fits
A written rubric is also what makes software useful in screening. Once criteria, anchors and sources are explicit, a system can compile the evidence for each assessment and flag what is missing, and a reviewer can check every line against its source. Without a rubric, any automated score is just another opinion nobody can audit.
When a fund wants its first-look research organized around its own criteria, we build fund-specific screening and research systems that work against the firm's rubric rather than a generic venture score. They separate source facts from inferences, keep analyst commentary and exceptions, and leave every decision about whether to advance a company with the investing team.