Evidence-based scoring rates a founder, company, candidate or deal on documented, checkable facts, with each part of the score linked to the record that supports it, so anyone reading the score can see why it was given and challenge the evidence rather than the reviewer.
How evidence-based scoring works
The rule is simple: no score without a receipt. Each criterion score carries four things:
- The claim. "Traction: 4 of 5."
- The evidence. "Annual recurring revenue of 1.1 million dollars, up from 400,000 dollars a year earlier."
- The source. Where the evidence came from: the data room, a filing, a reference call, a code repository, an interview answer.
- The date and status. When it was observed, and whether it was confirmed or is still the founder's or candidate's own claim.
When evidence is missing, the score says so rather than defaulting to average. "Unknown" becomes a question for the next meeting. Filling known gaps before scoring is the job of data enrichment, and when evidence arrives on different scales, such as revenue in two currencies, normalization makes it comparable first.
The term borrows from evidence-based management, which asks decision makers to combine evidence from scientific research, the organisation's own data, experienced professionals and affected stakeholders (CIPD, Evidence-based practice).
Why evidence-based scoring matters
Impressions are fast and persuasive, and they are hard to argue with because there is nothing to check. Evidence can be checked, so disagreements become productive: either the revenue figure is right or it is not.
There is also a long research record behind rating on recorded facts with a fixed method. Reviewing decades of studies, including meta-analyses by Grove and colleagues and by Kuncel and colleagues, Meijer and colleagues conclude that combining information with a fixed rule matches or outperforms combining it in the mind, yet practitioners seldom do it (Meijer et al., 2020).
For a venture fund, evidence-based scoring means a partner's "great founder" becomes "sold her last company, a freight billing tool, to a logistics group in 2023, and two former customers confirm she personally ran their onboarding". For a hiring team, it means every place on a ranked shortlist traces back to cited facts, which is also a defence when a rejected candidate asks why. For a PE firm, it makes the screening work reusable in diligence and the investment memo.
Worked example: one candidate, two scorecards
Candidate B is interviewing for a sales lead role at a fictional company, Oakline Health.
| Criterion | Impression-based | Evidence-based |
|---|---|---|
| Closing ability | 5: "Clearly a closer" | 4: Closed 14 hospital contracts in two years, per the reference from her former manager; average deal size not confirmed |
| Domain knowledge | 4: "Knows healthcare" | 3: Sold to clinics, not hospital procurement; answered the structured question on procurement cycles at a 3 level |
| Team leadership | 4: "Natural leader" | Unknown: has not managed a team; ask in the next round |
The impression-based version averages 4.3 and looks like a strong hire. The evidence-based version is lower, has a gap that needs filling, and tells the next interviewer exactly what to ask. Each line on the right is a signal someone else can check; it is also the shape a good scorecard takes.
Evidence-based scoring vs data-driven decisions vs gut feel
| Evidence-based scoring | Data-driven decision making | Gut feel | |
|---|---|---|---|
| Inputs | Facts, including qualitative ones, each cited | Mostly quantitative data | Impressions |
| Handles interviews and references | Yes, recorded and cited | Often ignores them | Yes, unrecorded |
| Traceable | Every point | At the dataset level | No |
| Main risk | Slower to record | Ignores what cannot be counted | Inconsistent and hard to review |
The scoring model sets the criteria and weights. Evidence-based scoring governs what each criterion score may rest on. Evidence-based does not mean numeric. A reference call or an interview answer is evidence if it is recorded and cited. That is how qualitative scoring stays rigorous.
Common mistakes
- Citing the deck as the source. A founder's slide is a claim. Note it as unconfirmed until checked.
- Evidence collected after the decision. Gathering facts to support a call already made is documentation, not scoring.
- Dropping what cannot be counted. Interview answers and references are evidence when written down.
- No trail. If nobody can reconstruct why a score was given six months later, the process fails on auditability, and for automated scoring, on explainability.
How ScoringFactory approaches evidence-based scoring
This is the core of what ScoringFactory does. It learns the bar from a team's past yes and no decisions, applies it to every founder, company or candidate, and cites each score to the line in the record behind it. Nothing is scored without a receipt, and your team makes the call. See how it works.
Frequently asked questions
What is evidence-based scoring?
It is a way of rating people, companies or deals in which every score is tied to specific, checkable evidence and its source. Instead of "strong team: 5", the record shows which facts earned the 5 and where they came from. Missing evidence is marked as unknown rather than guessed.
How do you link a score to its evidence?
Add an evidence field to every criterion on the scorecard and require it before the score can be submitted. Record the fact, its source, the date, and whether it has been confirmed. Software can attach the source document or line directly. A score whose evidence field is empty should be treated as unknown.
Is evidence-based scoring the same as data-driven decision making?
They overlap but differ. Data-driven decision making usually means relying on quantitative data. Evidence-based scoring includes qualitative evidence, such as interview answers and references, as long as it is recorded and cited, and it ties evidence to each individual score rather than to a decision as a whole.
Does evidence-based scoring remove bias?
It reduces some bias by making reasons visible, so a score based on an impression stands out. It does not remove bias in the evidence itself: if a criterion rewards something unevenly available, like a particular employer, recording it carefully does not make it fair. That still needs checking across groups.
Sources
- Evidence-based practice for effective decision-making, CIPD, 2023
- A tutorial on mechanical decision-making for personnel and educational selection (Meijer, Neumann, Hemker and Niessen, 2020), Frontiers in Psychology, via PubMed Central