Skip to main content
Blog

Decision-making

How to score a startup without fooling yourself

Keep fit and strength apart, give your fund's weights a number of their own, leave out what you don't know, and let one deal-breaker hold the score down. Five rules, and how Aventro's score follows them.

Dor Tagger

Co-founder & CTO, Aventro

November 20, 2026

4 min read

Short answer: a startup score is worth using when you can say what it means and what it left out. Five rules get you most of the way. Score fit and strength separately. Give your fund's weights a number of their own. Weigh by stage and by kind of company. Leave out what you don't know. And let a single deal-breaker hold the score down.

Most funds have a scorecard somewhere. Team out of 10, market out of 10, traction out of 10, average it. The trouble starts when the number travels. It goes into the pipeline view, then the IC deck, and by then nobody remembers that the 7 for traction came from the founder's own spreadsheet.

These are the rules we built Aventro's score on. We got most of them wrong first.

1. Fit and strength are two questions

A great company outside your mandate and a weak one inside it can land on the same blended number. That's the worst thing a score can do: two opposite situations end up looking the same.

So keep two numbers. Fit asks whether it's your kind of deal: stage, sector, cheque size, geography, the things you never do. Strength asks whether it's a good company.

Treat your exclusions as gates. A company in a sector you never back shouldn't lose a few points and get through on a strong team. In Aventro an exclusion is checked in code against the company's facts, and a breach comes with the reason a partner reads. When the facts can't clear it, because the sector or the stage is missing, a person gets the question.

2. Your weights get their own number

Funds weigh things differently, and they should. A deep-tech fund cares more about IP than a consumer fund does.

The mistake is letting those weights change the company's score. Then the same company reads 71 at one fund and 58 at another under the same name, and the number stops meaning anything outside your office.

We used to do exactly that. Now the Aventro Score is the same at every fund, weighed by the kind of company and its stage. A fund's own levels make a second number, the Fund Score, from the same factor scores. Change your levels and every Fund Score is re-weighed by plain arithmetic. No new analysis, and it can't drift from the evidence the Aventro Score stands on.

3. Weigh by stage and by kind of company

A pre-product company has no traction to score. Weighing it like a Series B company punishes it for being early.

In Aventro's score the team carries 40% of the weight at pre-product. At the early stage traction gets 15% of its own and the team drops to 25%.

The kind of company matters just as much. Our weights were first calibrated for software, so a preclinical drug company was scored partly on revenue it was years from having and never on its regulatory pathway. A drug company now has weights of its own: at pre-product, IP carries 22% and the regulatory route 12%. The software weights give the regulatory module nothing at that stage.

4. Leave out what you don't know

If the founder left unit economics blank, the honest score for unit economics is "unknown". Scoring it zero punishes the company for something you didn't check. Scoring it a five makes something up.

Leave it out and let it lower your confidence instead. In Aventro a factor is only scored when its inputs exist, and each area is averaged over the factors that are there. Unit economics before real revenue are projections, so they go to judgment instead of a formula.

Then make confidence mean something. Ours used to be the number of modules that ran, divided by ten. It read 100% on every complete analysis, whether the data room held forty documents or none. Now it's the share of the score backed by something other than the company's own word: a document, a check against an independent source, a record found outside the company. A number the founder typed into a form still counts as their word.

A low figure means a thin file. At seed that's normal, and it's exactly what a partner should know before the meeting.

5. One deal-breaker shouldn't be averaged away

Averages are polite. If a company has no founding team, a strong market score shouldn't carry it back up to a respectable number.

We found this on a company we ran through Aventro ourselves. A deal-breaker was triggered and the score still read 47, above companies with nothing disqualifying at all. Now, while a deal-breaker is open, Aventro holds the score at the floor where a company reads as not investable, and every place that shows the number says it's held, and by what.

Before you trust a score

Ask any score three things, ours included:

  • What would move it, and by how much?
  • How much of it rests on the founder's own word?
  • What did it leave out?

In Aventro, What-If scenarios answer the first one. A score is decision support. It tells you where to look, and your partners still make the call. The full method, factor by factor, is on the scoring page.