How to Set Up a Startup Selection Committee

A selection committee is the group that makes the final call on which candidates a program backs. Set up well, it decides quickly and can defend every choice. Set up badly, it becomes five opinions that cancel each other out, and the decision defaults to whoever argues hardest. This guide covers how to compose the committee, define its roles, give it a common basis, and run a session that produces a defensible decision instead of a stalemate.

Most programs assemble a committee by gathering smart people in a room and hoping consensus emerges. It rarely does. Without structure, the discussion drifts to whichever company had the most memorable pitch, the loudest advocate wins, and nobody can reconstruct afterward why one candidate made it and another did not. The problem is almost never the quality of the people. It is the absence of a shared basis and clear roles.

This guide walks through the pieces that make a committee work: who is on it and why, the roles that keep a session moving, the common basis every member scores against, how to run the meeting itself, how to handle conflicts of interest, and how to leave a record that defends the outcome later.

Who should be on it, and how many

The first mistake is size. A committee of twelve is not more rigorous than a committee of five, it is slower and less accountable, because responsibility diffuses until no one owns the decision. For most programs, three to five voting members is the range where every voice is heard and every member feels the weight of the call.

Composition matters more than headcount. A committee that is all one profile, all investors, all technical, all insiders, will share the same blind spots. The useful mix pairs people who see different things: someone who reads teams well, someone who reads markets, someone with domain depth in what you back, and someone whose job is to be skeptical rather than enthusiastic. The goal is not balance for its own sake. It is to make sure the thing one member would miss is the thing another is watching for.

Define the roles before the first meeting

A committee without roles is a conversation. A committee with roles is a process. Four roles cover most programs, and one person can hold more than one in a small group:

• The chair. Runs the session, keeps time, and makes sure every candidate gets the same depth of discussion rather than the first three getting an hour and the last ten getting a shrug. The chair does not decide; they make the deciding orderly.
• The reviewers. The voting members who score and argue the candidates. Their job is to apply the common basis, not to advocate for favorites.
• The domain expert. Brought in for depth on the specific area a cohort covers, sometimes as a non-voting advisor. They answer "is this technically real" so the reviewers do not have to guess.
• The recorder. Captures what was decided and why, in a form that survives the meeting. This is the role programs skip and regret, because a decision with no record cannot be defended six months later.

Give the committee one basis to score against

This is the step that separates a committee from a debate club. If each member weighs their own criteria, two candidates are never actually compared, and the ranking is an artifact of who reviewed what. A common basis fixes that: a set of criteria, decided in advance, applied identically to every candidate.

The Deckwise Method frames the basis as three questions kept deliberately apart: is it true, meaning do the material claims hold up against evidence; is it a good company, objectively; and is it a good fit for this program specifically. Keeping them separate matters because a company can pass one and fail another, and a committee that fuses them into a single impression loses exactly the distinctions it exists to weigh. The committee should also see, before it meets, which claims were verified and which could not be, so the discussion is about a checked candidate rather than a polished narrative.

The weighting between these questions is set once, for the program, and applied to all. A committee that re-tunes its criteria per candidate is not comparing candidates, it is rationalizing preferences.

Run the session so it decides

A good session has a shape. The candidates arrive already scored on the common basis and already narrowed to a shortlist, so the committee spends its scarce time on the decision, not on triage. For each candidate, the chair walks the committee through the score, the verified claims, and the flags, then opens a bounded discussion, then calls the vote. Bounded is the key word: an open-ended debate expands to fill any amount of time and rewards stamina over judgment.

Two mechanisms keep the ranking honest inside the room. A wildcard surfaces a candidate that is average overall but exceptional on one dimension, so a standout is not lost to a middling total. A red flag pulls a score down on purpose, so a serious negative cannot hide behind strength elsewhere and get waved through. Both exist so that the vote reflects what the committee actually believes, not what the arithmetic happened to produce.

Handle conflicts of interest explicitly

Every committee eventually faces a candidate a member knows: a former colleague, a portfolio company, a friend. Handled quietly, this poisons the decision and the record. Handled explicitly, it is routine. The rule is simple and written down before it is needed: a member with a material connection to a candidate declares it, and either recuses from that candidate's vote or has the connection noted alongside their score. What matters is that the conflict is visible, so the decision can be defended and no one has to wonder later whether a vote was really about the company.

Leave a record that defends the decision

The output of a committee session is not just a list of yeses and noes. It is a record of why: which claims held up, how each candidate scored on the three questions, what was flagged, who recused, and the reasoning behind each call. That record is what lets you answer "why this one and not that one" to a board, an LP, or an unsuccessful applicant, months after the memory has faded.

This is also where the committee's work meets its obligations. Where a program uses software to help rank candidates, human oversight, transparency, and traceability are expected, and a committee that decides on a visible basis and records its reasoning is already meeting that standard. The record is not bureaucracy. It is the difference between a decision you made and a decision you can stand behind.

Frequently asked questions

How many people should be on a selection committee? For most programs, three to five voting members. Larger committees are slower and diffuse accountability without adding rigor. Composition matters more than size: mix people who read teams, markets, and technical depth, plus someone whose role is to be skeptical.

What is the difference between a reviewer and the chair? Reviewers score and argue the candidates on the common basis. The chair runs the session, keeps every candidate to equal depth, and calls the votes, but does not decide the outcome. The chair makes the deciding orderly; the reviewers do the deciding.

How do you keep a committee from just picking the best pitch? Give it a common basis decided in advance and applied to every candidate, bring candidates in already scored and verified, and bound the discussion. A wildcard protects one-dimension standouts and a red flag stops a serious negative from hiding behind a strong average.

How should conflicts of interest be handled? With a written rule applied before it is needed: a member with a material connection to a candidate declares it and either recuses from that vote or has the connection recorded next to their score. The point is that the conflict is visible in the record.

Why does the committee need a written record? Because "why this one and not that one" gets asked months later, by boards, LPs, or applicants, and a decision with no record cannot be defended. A record of the basis, scores, flags, and reasoning is also what meets the transparency and traceability expected when software helps rank people.

The bottom line

A selection committee works when it is a process, not a conversation: three to five members with complementary blind spots, clear roles, one basis applied to every candidate, a bounded session that decides on a pre-verified shortlist, conflicts declared openly, and a record that explains every call. Assemble it that way and the committee decides quickly and defends its choices. Assemble it as a room full of opinions and it will default to the loudest one.

How this shows up in Deckwise

Deckwise brings the committee a shortlist that is already scored on one basis and already verified, so the session is spent deciding rather than triaging. Every candidate arrives with its score across the three questions, the verified claims and their sources, wildcards for one-dimension standouts, and red flags that pull serious negatives down where the committee can see them. Evaluators are included rather than charged per seat, so involving the whole committee costs nothing, and the reasoning is exportable as the record that defends the decision. The committee always makes the call.

Related: The Complete Guide to Startup Screening · What Is a Defensible Shortlist? · The Deckwise Method