A rubric is a written guide, agreed before any review starts, that lists the criteria being judged and describes what the evidence looks like at each score level, so two reviewers rating the same founder, company or candidate land on the same number.
What a rubric contains
A rubric is a grid. Rows are criteria. Columns are score levels. Each cell holds a short description of the evidence that earns that score. Rubrics are standard in education, where teachers use them to grade student work consistently, and a 2020 review by Panadero and Jonsson found that most published criticism of rubrics rests on anecdote rather than direct evidence, while empirical support for their benefits keeps growing. The same tool works for deals, hires and vendor scoring.
There are two common shapes:
- Analytic rubrics score each criterion separately, then combine them. Better for comparing many candidates or companies, because you can see where each one is strong or weak.
- Holistic rubrics give one overall score with a description per level. Faster, but two people can reach the same 3 for different reasons.
The strongest rubrics use behavioral anchors: each level is described with an observable example instead of an adjective. "Has shipped a product used by paying customers" is an anchor. "Strong execution" is not. This approach has a name in HR research: behaviorally anchored rating scales. SHRM's guidance on interview rating scales recommends the same thing for hiring: anchor each level to observable, job-related behavior, and train raters on examples of strong and weak answers. Anchors matter most for soft criteria such as clarity or ownership, which is the territory of qualitative scoring.
Why rubrics matter for investors and hiring teams
Without a rubric, a 4 out of 5 records how the reviewer felt, and feelings differ. With one, the score records which description the evidence matched. That is what makes scores from different partners, interviewers or quarters comparable, and it is the first thing to fix when inter-rater reliability is low.
The U.S. Office of Personnel Management lists it as a defining feature of a structured interview: every response is evaluated against the same rating scale and the same standards for an acceptable answer (OPM, Structured Interviews). A venture fund gets the same benefit at first meetings: the partner who met the founder and the partner who only read the notes can argue about the evidence instead of the adjective.
Example: one rubric row, written three ways
Take a single criterion a seed fund might use, customer evidence. Here is how the same row improves as the anchors get more concrete.
Weak (adjectives only): 1 = poor, 3 = good, 5 = excellent.
Better (levels described): 1 = no customers, 3 = some customers, 5 = many customers.
Anchored:
- No conversations with buyers on record.
- Interviews with potential buyers, no commitments.
- At least one signed pilot or letter of intent.
- Paying customers, fewer than ten.
- Ten or more paying customers, with at least one renewal or expansion.
Give the anchored version to two partners reviewing a fictional company, Ledgerly, with four paying design partners and no renewals yet. Both should score it a 4. With the first version, one might say 3 and the other 5.
Rubric vs scorecard vs evaluation rubric
These three words get used interchangeably. They are different documents.
| Rubric | Evaluation rubric | Scorecard | |
|---|---|---|---|
| What it is | The general tool: criteria and level descriptions | A rubric built for one decision type, with weights | A rubric filled in for one person or company |
| Written when | Before any review | Before review, per role or fund strategy | During or after each review |
| Example | A 1 to 5 scale for customer evidence | The full seed-stage rubric for a fintech fund | Ledgerly's scores and notes from the first meeting |
Put simply: the rubric is the answer key, the scorecard is the marked paper.
How to write a rubric
- List the four to seven things that actually separate good outcomes from bad ones. Drop anything you would never reject on.
- For each criterion, write the top and bottom anchor first. The middle is easier once the ends are fixed.
- Use evidence a reviewer can point at: a number, an event, a document, an answer to a set question.
- Decide how much each criterion counts. See weighting. Criteria, anchors and weights together make a scoring model.
- Have three people score the same five past cases. Where they disagree, rewrite the anchor, then run a calibration session.
Common rubric mistakes
- Anchors that are adjectives. "Exceptional" means something different to every reader.
- Two ideas in one row. "Technical depth and communication" forces a reviewer to average two different things.
- A rubric written after the review. At that point it tends to describe the favourite candidate.
- Never revisiting it. If every company scores 3 or 4 on a row, the anchors are not separating anyone.
How ScoringFactory uses rubrics
Most teams have a rubric in their heads that never made it onto paper. ScoringFactory learns it from the yes and no decisions a team has already made, then shows each score next to the record it came from, so reviewers can check the reasoning row by row. See the flow.
Frequently asked questions
What is a scoring rubric?
A scoring rubric is a table of criteria and score levels with a short description of what the evidence looks like at each level. Reviewers match what they see to a description rather than picking a number by feel, which makes scores from different people and different weeks comparable.
How do you write a rubric for evaluating candidates or deals?
Choose a handful of criteria that predict success, write concrete anchors for the top and bottom of each scale, fill in the middle, and test it by having several people score the same past cases. Rewrite any row where they disagree. For a full worked version for one role or fund, see evaluation rubric.
What is the difference between a rubric and a scorecard?
A rubric is the blank standard, written once and used for everyone. A scorecard is that rubric filled in for one candidate or company, with a score and supporting evidence for each criterion. One rubric produces many scorecards, and because they share the same anchors, those scorecards can be laid side by side and compared row by row.
How many score levels should a rubric have?
Four or five is common. Five gives a natural middle, which some teams want and others avoid because reviewers park uncertain cases there. Four forces a lean either way. Fewer than three loses detail, and more than seven is hard to anchor with distinct, observable descriptions.
Sources
- Structured Interviews, U.S. Office of Personnel Management, current text
- Panadero and Jonsson (2020), A critical review of the arguments against the use of rubrics, Educational Research Review (Elsevier)
- Build Consistent Hiring Decisions with Job Interview Rating Scales, SHRM, 2026