Stores
Stores
—
Opens this month
—
Sizings this month
—
Today
—
measured from a photo
All time
—
Shoppers
—
All stores
Loading…
Widget health, last 7 days
Loading…
Analytics
Sizings over time
Loading…
Store leaderboard
Busiest stores. Click one to scope everything to it. Counts sizings and size checks together, as activity.
Loading…
Top garments
Most-opened products across the platform. Includes size checks, so this reads as shopper interest, not measurement volume.
Loading…
Size mix
Recommended sizes across the run. Photo sizings only. A size check re-shows the same shopper on another product.
Loading…
Waist-read confidence
Low confidence usually means a loose garment hid the waist, which is a capture-quality signal.
Loading…
Weight given
Typed weight is trusted and sharpens the fit; estimated is the model's guess.
Loading…
Based on up to the 500 most recent events in range. "Sizings" are measured from a photo; "size checks" are a returning shopper being re-shown their saved size on another product (free, and logged the moment they open the widget). Hip/chest, confidence and weight exist on photo sizings only. "Try-ons" counts rendered garments, which attach to an existing sizing event rather than logging their own, so a size check that led to a try-on was real engagement rather than a bounce. "Added to cart" is the shopper tapping through on the size we gave them. That is intent rather than revenue, since the cart can still be abandoned.
Platform-wide over the window above. Fallback rate is deliveries that needed a second backend; p50/p95 are the shopper's whole wait on a delivered preview.
Loading…
Try-on reliability
Every render attempt, not just the ones that reached a shopper. A rising "failed" share is a backend problem worth a look before it's a pattern.
Loading…
By backend
Attempts and average latency per backend. The number to watch before flipping
TRYON_BACKEND.Loading…
Try-on spend
Loading…
Spend per day
Loading…
FLUX prices are what OpenRouter actually charged, recorded from 25 Sep 2026. Gemini prices are our estimate from
token counts, because Google doesn't give a price per picture. Older FLUX prices are estimates too, and they priced
only the finished picture, not the photos sent in, so they read low. In the one try-on we measured, the old
estimate was about half of what OpenRouter charged. "Per picture a shopper saw" divides all
spend, failed attempts included, by the pictures actually delivered.
Usage & free tiers
Loading…
Numbers are collected periodically by
scripts/collect-costs.mjs; run it to refresh.