Every channel you pay grades its own homework, and every dashboard flatters its owner. This page is the full argument for the one number that can't lie — revenue measured against a control group — and how LiftSignals makes it automatic on every send.
Maya buys an $80 jar of night cream. Your email platform claims the sale — she opened Tuesday's campaign. Meta claims it too — she scrolled past an ad last week. Search claims it as well — she Googled you before checkout.
Add up the dashboards and you apparently made $240 on an $80 order. Every tool is telling the truth about what it touched. None of them are telling you what caused the sale.
Even if only one tool claimed the sale, there's a deeper problem: your best customers — the exact people your campaigns target — are the people most likely to buy with no campaign at all.
Email thirty loyal customers, eight buy, and the dashboard credits all eight to the email. But how many of those eight were coming back regardless? Without a control group, that question is unanswerable. With one, it's arithmetic.
At push time, 3–33% of the audience is withheld as the control — never fewer than 100 people, so noise can't masquerade as signal.
Group membership and the measurement window lock in the same transaction as the send. Nothing can be gamed retroactively — by anyone, including us.
For 21–45 days depending on campaign type, orders from contacted and held-out customers are tracked identically, in integer cents.
Lift = what the contacted spent minus what the holdout spent anyway. It's marked proven only when the evidence clears the statistical bar.
The Lift panel always shows the strongest evidence you've actually earned — and says plainly which kind it is. New workspace? It says measuring, with directional signals. Experiments open? It says tracking, with a date. Only completed holdout tests get the word proven.
That restraint is the point. A number that refuses to exaggerate on the way up is a number your CFO can trust at the top.
The holdout is small — as little as 3% on large audiences — and it's the price of knowing whether the other 97% did anything at all. One unmeasured campaign that "would have happened anyway" costs far more than a control group ever will. And customers in a holdout aren't lost; they're just not in this send.
Then LiftSignals says so instead of pretending. There's a statistical floor — audiences too small for a meaningful holdout are handled honestly, and the panel shows directional evidence rather than a fabricated "verified" number. When it says proven, the math is actually there.
Everything else — the briefing, the audiences, the closed-loop wins ("flagged yesterday, bought today") that land within days. The Lift panel shows "tracking" with a date while windows are open, and campaigns keep launching in parallel. The proof lags the work by design; it never blocks it.
Connect a store — or a spreadsheet — push one audience, and let the control group do the talking. Free for 30 days — no credit card required.