How does cofounder matching actually work?
Trusted Cofounder reads each brief as plain text, parses it into structured signals (role, skills, stage, location, commitment, behavioral fit) and ranks every opposite-role candidate against the brief on a 0-100 score. The five dimensions are weighted 40 / 25 / 15 / 10 / 10. No embeddings, no black box.
What is a brief?
A brief is a short, first-person description of what you need or what you want to join. Founders write briefs describing the kind of cofounder they are looking for. Cofounders write briefs describing the kind of company they want to help build. There is no form: the composer is a single textarea with a working-style sub-block.
A parser reads the rawText and extracts structured signals: the roles you need, specific skills, your stage on the founder ladder (exploring, have-idea, building-mvp, post-launch), your commitment band (full-time, side-project), your city or remote-ok flag, and the industries you care about. Optional: a behavioral block that captures how you work.
A brief cannot go live until it has at least 80 characters of raw text, a stage, a role or skill list, a commitment, and either a city or remote-ok. The gate exists so the matching pool stays useful: a stale or empty brief pollutes every viewer's ranked results.
What does the match score mean?
The score is the sum of five dimension scores, each weighted by its relative importance. The weights are fixed and visible:
| Dimension | Weight | What it measures |
|---|---|---|
| Role coverage | 40% | Whether the candidate's skills and tags cover the roles and skills the brief asks for. |
| Stage and commitment fit | 25% | How close the candidate's stage readiness and full-time / side-project signal are to what the brief needs. |
| Location | 15% | Same city scores 100, remote-ok match 70, same country 40, mismatch 0. |
| Industry affinity | 10% | Whether the candidate's past work overlaps the brief's industries. |
| Behavioral fit | 10% | Compatibility on a small set of behavioral axes derived from each side's fingerprint. |
The breakdown is visible on every match card. Click the score and you see how each dimension contributed.
What gets filtered out before scoring?
Two hard filters drop candidates from the result list rather than merely lowering their score. First, opposite role: a founder's brief only ranks cofounders, never other founders. Second, commitment mismatch at the band level: a full-time brief never ranks a side-project candidate, and vice versa.
Soft mismatches (e.g. a founder building MVP looking for a post-launch cofounder) score low but stay visible. The scoring logic does not pretend mid-stage gaps disqualify the relationship.
How does behavioral fingerprint contribute?
When a user connects sources like GitHub, Stack Overflow, a blog, or Semantic Scholar, an enrichment pipeline builds a behavioral fingerprint: a short summary, key strengths, and a set of behavioral axes (e.g. risk appetite, depth vs. breadth bias). The fingerprint earns the Verified Builder badge and contributes 10 percent to the match score when both sides have one.
The fingerprint is bonus credibility, not the matching hero. Briefs and the deterministic role/stage/commitment dimensions do most of the ranking work. When either side has no fingerprint, the 10 percent is redistributed across the remaining dimensions, so a brand-new account is not penalised for the absence of an enriched source.
Why deterministic, not embeddings?
Two reasons. First, transparency. A 40 / 25 / 15 / 10 / 10 breakdown is auditable; an embedding-similarity score is not. A founder reading “Why is this candidate ranked third?” deserves a sentence-level answer, not a vector distance.
Second, scale. At MVP scale (under 5,000 eligible users) brute-force candidate fetch plus in-process ranking is cheap, predictable, and easy to debug. Embeddings would add an entire second system, and the gain in ranking quality at this scale is not measurable.
What does the engine do that a profile-first directory cannot?
Profile-first directories rank candidates by category buckets (16-skill picklists, 20-field structured forms). The brief-first approach reads “I am building a vertical SaaS for Finnish construction companies, post-launch, full-time, Helsinki” and ranks against that exact sentence. The candidate ranked first is the one whose profile actually overlaps that exact sentence, not the one who ticked “Software Development” and “B2B SaaS”.
The cost is that you write a brief instead of clicking dropdowns. That cost is the feature. The brief is what makes every match useful.
Where does this break?
- Briefs that are template-shaped get archived by the nightly brief-quality sweep. The hint surfaces in the composer before publish.
- Briefs that don't name a stage or commitment can't publish. The publish gate is hard.
- A user with no active brief is not in the pool, regardless of profile completeness. Pool quality is brief quality.
Read about finding a cofounder in Finland for the live numbers, or compare the platforms in this side-by-side.