Short answer: A good startup screening score compares each startup's materials (deck, one-pager, website) against one written investment thesis. It scores criteria such as sector and stage fit, team, market, traction, business model and risks, then explains each score in words. Automated screening is good at catching weak or off-thesis deals fast. It is not good at judging founders you haven't met, so use it to rank your pipeline, not to make the final call.

If you run an accelerator or a fund, you know the problem. 300 applications land in the last week, and someone has to read every deck. That's where screening tools come in.

The data that goes into a score

  1. Your thesis. The score is only as good as what you compare against. "B2B SaaS, pre-seed to seed, Europe, revenue or paid pilots" scores very differently from "deep tech, any stage".
  2. Sector and stage fit. Is this even in scope? This alone removes a big share of applications.
  3. Team. Founder backgrounds, relevant experience, technical depth, full-time or not.
  4. Market. Size, timing, and whether the deck shows a believable path to it.
  5. Traction. Revenue, pilots, users, growth rate, letters of intent.
  6. Business model. Who pays, how much, and how often.
  7. Risks and gaps. Missing financials, unclear ask, regulatory exposure, crowded space.

Each criterion gets a score and a short written reason. The reasons matter more than the number: they let your team disagree with the AI in seconds instead of re-reading the deck.

Does automated screening catch weak deals?

Where it works well

  • Off-thesis deals: wrong sector, stage or geography. Near-perfect at catching these.
  • Missing basics: no traction data, no clear ask, no team slide.
  • Consistency: every deck is judged on the same criteria, at 2am or 2pm.

Where it struggles

  • Founder quality that isn't on paper. A great founder with a weak deck can score low.
  • Non-obvious markets. Truly new categories don't match patterns well.
  • Claims it can't verify. The AI reads what the deck says; it can't confirm the revenue is real.

So the right workflow is: screen everything automatically, review the top 20 to 30% by hand, and do a quick human pass on low scorers with strong teams.

What to look for in a screening tool

  • Scores against your thesis, not a generic "good startup" model.
  • Written reasons for every score.
  • Bulk upload, so a whole cohort is screened in one go.
  • A way to export or share the ranked list with your committee.

How Gatsbie does it

Gatsbie's screening tool lets you upload many startups' decks at once, scores each one against your investment thesis and ranks them, with reasons. Accelerators like Nucleate use it to review cohorts. Read more in Gatsbie for investors: screen deal flow against your thesis, or see Gatsbie pricing for enterprise plans for accelerators and teams.

Written by Yen Tran, founder of Gatsbie.