Method
How the Leak Engineprices a leak.
Every figure Anry gives your agent is computed in code, from your own store’s data. Here’s what the engine reads, compares and won’t claim.
Rules
What the engine always does.
Same data, same figure.
Run it twice on the same data and you get the same findings and the same money.
Your checkout, not their claims.
Revenue comes from your store’s orders. What Meta or Google claim never decides a finding.
Judged against your own store.
Every page and ad is measured against your own store. Never against someone else’s.
Enough data first.
No verdict without enough visits. Until then your agent hears “too early”, not “all clear”.
Written once, never edited.
Every finding is dated and frozen. Each run writes a new report, so you see exactly what changed.
Numbers from the engine, words from a model.
A model writes the explanation and the fix, and checks if an ad’s offer matches its page. It never sets a figure.
Checks
What it compares.
On your site
Maxi shirt dress
$148
Exits, page by page
Drop-off
Visitors leave this page far more than the rest of your store. The ones who stay, buy.
Long wool coat
$289
Money lost a month, by page
Revenue leak
A normal-looking page that still loses more than your others.
TikTok adsWinter creators campaignLong wool coat
$289
Grey knit
$129
Wool coat
$249
Grey tee
$49
Bounces on /collections/winter
Traffic mismatch
One source bounces off a page that works for everyone else. The traffic is wrong, not the page.
Google searchSearched “warm gift ideas”
Cashmere crew
$189
Grey knit
$129
Grey tee
$49
Wool coat
$249
Revenue a visit, by source and page
Traffic opportunity
A source and page earning well above their peers. Scale it.
On your ads
The dress of the season is back.
Spend and sales, by hour and country
Where it runs
Every ad split by country, device and hour. The slices that spend without selling get a price.
Performance MaxWinter sale campaignSponsored
Winter style guide 2026
Layering, fabrics and the coats of the season. Free returns.
Sales a visit, by landing page
Where it lands
Every page and product your Google ads land on, judged by what visitors buy there.
Cashmere, softer than ever.
30% offCashmere crew
$189 · full price
Offer and page
Whether the ad’s offer matches the page it opens. A model answers yes or no; the engine sets the price.
The coat, now in grey.
Running since 14 AugEngagement, week by week
Creative wear
Engagement falling against the ad’s own earlier weeks.
Ads are priced against a target return you can set.
Limits
What it won’t claim.
- 1−Forecasts of what a fix will earn
- 2−Causal claims
- 3−Scores and probabilities
- 4−Benchmarks from other stores
- 5+Money already lost, in your own data
- 6+Evidence you can replay, visit by visit
- 7+Money lost as the only trigger
- 8+Your store as its own baseline
No forecasts.
Only money already lost. No guesses about what a fix will earn.
No causal claims.
A finding says where and how much. The why comes from the evidence, and you can replay it.
No scores.
Money lost is the only trigger. No scores, no probabilities.
What stays private. We publish what each check compares. Thresholds, weights and windows stay private, and they never change between runs.
Spec
Under the hood.
$ claude mcp add --transport http anry https://mcp.anry.io/mcp
✓ anry added · sign in when your browser opens
anry · 16 tools, all read-only · 3 prompts
› What’s costing my store the most?
● anry · get_revenue_leaks
⎿ /products/maxi-shirt-dress · $1,200–1,500 a month
See it on your own store.
Start a trial, add the tracker, connect your ads. Your ad history is read at once.
14-day free trial · No credit card required