Report · Ninety‑day pilot · Issue No.03

What thirty‑eight merchants saw.

Five percent of shoppers see nothing. The other ninety‑five see Osmo. The numbers below are the gap between them, checked against Shopify orders.

Contribution margin
14 %

median lift in margin per shopper across pilot stores.

Decision latency
38 ms

how long Osmo takes to decide, ninety‑nine percent of the time.

App count
5 →1

apps collapsed into a single engine.

Renewal
94 %

of pilot merchants renewed after ninety days.

Fig. 1 · Margin lift by vertical · n=38 merchants

Every vertical moved. Some moved more.

Stores with bigger catalogs and price‑sensitive shoppers see the largest lifts — the engine has more offers to choose from and more kinds of shoppers to tell apart. No vertical in the pilot regressed.

Apparel
+18.2%
Beauty
+15.4%
Home & goods
+13.5%
F&B · pantry
+12.1%
Consumer elec.
+10.2%
Outdoor
+8.4%
n = 38 · 90 days 5% holdout checked against Shopify orders
Fig. 2 · Cumulative margin lift · treatment vs. holdout

The lift shows up by week two.

For the first two weeks, Osmo is still figuring out which offers work for which kinds of shoppers. By day fourteen, the best offers pull ahead; by day thirty, the lift is real and statistically defensible on most stores. The five percent who saw nothing stay flat — that’s on purpose, so you can always tell what changed.

Day 14
offers separate
Day 30
lift is real
Day 60
settles · keeps measuring
Method · How we count

Margin, not conversion. Delta, not vanity.

We don't report conversion rate, because conversion rate is easy to move with a twenty percent blanket discount and nobody should celebrate that. We report margin per shopper, measured against a five percent permanent holdout that never sees any intervention. The delta is the only thing we pay ourselves on.

Primary
Margin per shopper
Secondary
90‑day margin / customer
Guardrail
Returns · same or lower
Next · Terms

One flat fee. A receipt that proves it.