Every number on this site traces back to this page. Nellie is paid a share of what she recovers, so the honest question is: recovered compared to what? Here is the exact answer, in the order a sceptical finance person would ask it.
If we cannot prove a sale would not have happened without her, we do not count it and we do not bill for it.
When Nellie starts at your store, we randomly withhold a slice of eligible shoppers – normally 10% – from every conversation she would otherwise have. They shop your store exactly as before. They never hear from her.
The split is drawn per shopper, not per cart, so nobody drifts between groups mid-relationship. It is fixed before she sends anything, and we cannot move someone after the fact.
Nellie texts them. Some reply, some buy, some ask her to stop.
Silence. Whatever they buy anyway is your baseline – the sales you were always going to get.
We compare revenue per eligible shopper in each group over the same window, then scale the difference across the treated group. The gap is incremental revenue – what only exists because she spoke to someone.
Deliberately boring arithmetic. No attribution windows, no last-click, no model deciding who deserves credit – just two groups and a subtraction.
Recovered revenue is net, not gross. Four deductions, applied before anything is invoiced:
A holdout only tells you something once enough people have passed through it. In a store's first weeks the gap between groups is often smaller than ordinary week-to-week noise, and we will say so rather than dress it up.
Until a result clears that bar, your statement shows it as provisional and we bill on the conservative end. We would rather under-claim early than publish a number that falls apart later.
Your monthly statement lists every conversation, which group each shopper was in, the orders on both sides, and the arithmetic above with your own figures. You can rebuild the total in a spreadsheet from your Shopify export – and we would like you to.