Apr 1, 20264 min read
Signal over noise: what a 70% match score actually means
Percent-match scores are usually marketing. Here's what ours computes, what it ignores, and where we draw the line on surfacing a score at all.
Every job site has a match score. Almost none of them explain it. We think that's a mistake — a score without a legend is theater, and theater erodes trust faster than no score at all. Here's what our match score actually is.
What the number counts
- Skill overlap between the role's extracted requirements and your profile — weighted by how rare each skill is across the index.
- Seniority match — junior/mid/senior bands, not exact title match.
- Salary intersection — your target band vs the listed band, with zero weight if the listing hides comp.
- Distance from your origin — soft penalty above ~30 miles, hard cutoff above ~90.
What it doesn't count
We don't weight company prestige, employer size, or funding stage. Those are signals you can apply yourself — sorting by them would launder our preferences into your search. We also don't factor in "similar people applied" collaborative-filter signals, because they reinforce existing networks instead of expanding them.
Why we gate on ≥60%
Below 60, the signal is statistical noise. Surfacing it anyway would give visual weight to jobs that don't deserve attention, and implicitly suggest that every role has a meaningful match to every user. Cards below the threshold simply omit the pill. The filter sidebar still lets you sort by match if you want the long tail, but cards don't yell it at you.