How Pulse scores work
A pulse score is an audit of public opinion — what reviewers, retailers' customers, and owners collectively think of a shoe. Here is exactly how we compute it, including the parts that are hard.
The problem with shoe ratings
Published shoe ratings cluster absurdly high. Across the 245 source ratings currently behind this site, the average is 8.7/10, and 13% are a perfect 10. Almost nothing scores below 8. When the median shoe scores 9, a "9.1/10" tells you nearly nothing — every shoe looks excellent, so no shoe does.
That's why our headline number is a percentile, not a rating. "93rd percentile among running shoes" is a claim about standing within a field. It can't be inflated by everyone being generous, because it measures shoes against each other rather than against an imaginary scale.
Where the numbers come from
Every source is typed by what it actually is, because a hands-on review and a star widget are different kinds of evidence:
- Independent reviews — outlets like RunRepeat or Running Shoes Guru that test shoes and publish scores.
- Retailer customers — aggregate ratings from stores like REI, Amazon, or Running Warehouse.
- Brand-site customers — owner reviews hosted on the manufacturer's own site. We include these: they're real customer sentiment, and the big brands' widgets track neutral retailers closely. But they're typed and weighted, because some brands' review displays run measurably hotter than the same data on neutral ground.
Sources live in a registry with a quality tier, and every rating we collect records where it came from, when we retrieved it, and what it said at the time. Sources move, change, and disappear — a dated observation stays honest even after its page is gone.
Correcting for each source's generosity
Sources grade differently. RunRepeat's ratings in our corpus average 8.9; Running Shoes Guru's average 7.9. An 8.5 from each is not the same signal — one is slightly below that source's normal, the other well above it. So before averaging, every rating is re-centred against its own source's historical average. You can see this on any shoe page under "How this score was built" — the arrow (9.4 → 9.6) is that correction, shown rather than hidden.
Each rating is then weighted by its source's tier and type, and the weighted average becomes the shoe's raw pulse. Finally the shoe is ranked against every scored shoe in its category to produce the percentile.
What the badge states mean
Ranked. The shoe's category has enough scored shoes (37 in running) for a percentile to mean something.
Scored, not ranked. We have the sources, but too few shoes in this category to rank against fairly. We show the raw score and coverage, and make no comparative claim until the field fills in.
Not yet scored. No usable ratings yet. We'd rather show nothing than a number we can't defend.
We also publish coverage — "7 of 8 sources" — because a score built from seven strong sources and one built from three blogs are different claims, and hiding that difference behind one number would be dishonest. New releases naturally start with low coverage.
What we get paid, and what it can't touch
Some links on this site are affiliate links; if you buy through one we may earn a commission at no cost to you. Scores are computed mechanically from the source registry by a script — commissions have no input into the arithmetic, and the arithmetic is displayed on every shoe page.
Current corpus: 55 scored shoes, 245 ratings across 61 sources. Re-scored on a rolling schedule; each shoe page lists its own sources.