A verdict, not a metric
Plain language: what is leaking, how much it costs a month, and what to do. Not a dashboard you have to interpret.
Money Leak
The campaigns that look like winners.
“Where am I losing money without knowing it?”
A leak is rarely an obviously bad campaign – those get paused in week one. A leak is the campaign reporting ROAS 2.8 whose customers never come back, or the keyword quietly buying people who were already typing your name. Both look healthy in the platform. Money Leak reads them against profit and repeat behaviour, and returns a verdict with its reasoning attached.
Open Money Leak in the demo →Sample data · no signup
What you actually see
The evidence underneath
Every verdict cites the spend, the matched orders, the margin and the repeat rate it was based on. Disagree with the reasoning rather than trusting a black box.
Lifetime value, not the first order
A campaign that fills the feed with one-time buyers stops looking like a winner by week six. First-order ROAS cannot see that; cohorts can.
A fix you can approve
Each verdict carries the specific change – pause this campaign, cut this keyword – ready to execute the moment you say so, and never before.
How it gets there
- 1
Match orders to ad clicks
First-party click-id capture ties an order to the click behind it. What cannot be matched is reported as unattributed rather than assigned to whoever claims it loudest.
- 2
Judge on money kept
Contribution margin, refunds and fees come off first. A campaign is measured on what it leaves behind, not on what it grossed.
- 3
Follow the customers forward
Cohorts track what an ad's customers spend over the following months, so acquisition quality shows up before the trend does.
Where it stops
What Money Leak will not tell you
- Attribution is imperfect. A customer who touches three channels before buying cannot be assigned to one with certainty, and we say so rather than picking one.
- A verdict is analysis, not advice. It tells you what the numbers say; the decision and its consequences remain yours.
- Cohort verdicts need history. In the first weeks after connecting, lifetime-value reasoning is thin by definition.
▸How is this different from the platform's own recommendations?
Meta and Google optimise toward the goal you gave them, measured with their own attribution. Neither knows your cost of goods, your refund rate or whether the customer came back. We do, and that is usually where the disagreement lives.
▸Will it just tell me to pause everything?
No. Roughly as often it says a channel has headroom and you are under-spending. A tool that only ever recommends cutting is a tool that has stopped measuring.
▸Can I see why it decided something?
That is the point. Every verdict shows the figures it used, and you can open the same numbers in Profit Tree and check them yourself.
▸What counts as wasted ad spend?
Spend that does not return its cost in contribution profit over the customer's realistic lifetime. That includes campaigns below break-even ROAS, keywords buying brand searches you already win organically, and audiences full of one-time buyers – all of which can look healthy in the platform's own reporting.
▸How is this different from just setting a ROAS target?
A ROAS target treats every product and audience as identical. A campaign selling 70%-margin goods breaks even at 1.4; one selling 30%-margin goods needs 3.3. Verdicts are computed against each campaign's actual economics, so the target is derived rather than guessed.
The module next to it
All six modules run on sample data in the open demo – no signup. Open Kepra →