Skip to content
KameleonAI
How we measure

How we count a sale

Every number we publish follows the same rules. Here they are, so you can hold us to them.

Every add to cart lands in one of three groups

Direct · counts

It put the product in the cart

Recorded at the moment it happens, never guessed. A tap on a product card it showed counts too.

Influenced · counts

The shopper added what it suggested

That exact product, after it recommended, compared or discussed it. Each add is graded, with the reason written down.

Organic · doesn't count

Everything else

Including any add we haven't graded yet. That one counts against us, so every figure is on the low side.

Who we compare against

Shoppers who talked to it, against shoppers on the same store it messaged who stayed silent. Same store, same weeks.

Talked to it6 in 100
Bedding store, 31 days: 6.06% bought
Stayed silent1 in 100
Same store, same weeks: 0.93% bought

Shoppers who choose to talk may already be keener to buy, so we say "shoppers who talked to it bought more", never "it caused".

Why we give a range

The low end assumes every order we couldn't trace came from a shopper who didn't talk to it.

The high end uses only the orders we could trace to a shopper.

When a number goes public

28+

Days per study

Anything shorter stays internal.

7

Days to settle

Every order in the window gets a week to land before we count it.

1

Store at a time

Each store on its own. Never pooled into one average.

0

Names without permission

A store is named by what it sells unless it says yes.

Your dashboard counts the same way