It reads your AdMob reporting data with read-only access, runs ten deterministic checks (geo floors, missing formats, demand concentration, eCPM cliffs, show/match leaks, consent gaps, serving restrictions, dead placements, config gaps, release regressions), and produces a report: how much money is leaking, the evidence, and step-by-step fix orders. It can then watch your account daily and verify whether applied fixes worked.

Click “Sign in with Google — read-only” on the home page, approve the consent screen (it requests only the admob.readonly scope), then click “Run the audit on my account”. The audit takes under a minute and ends at a shareable report link.

Synthetic demo accounts, each engineered with one classic leak (no floors, missing rewarded, single network, eCPM crash, placement bug, broken consent, ad-serving limit, broken release). They run through the exact same audit engine as real accounts — use them to see what a report looks like before connecting anything.

About 14 days of live traffic for a meaningful audit; 56 days unlocks the trend and release-regression checks fully. Accounts with no recent activity get a friendly “nothing to audit yet” state — that's expected, not an error.

Read-only reporting access (the admob.readonly OAuth scope) plus your basic Google profile (name, email). Florin cannot change floors, ad units, mediation groups, or anything else — every fix is applied by you.

Go to myaccount.google.com/permissions, find Florin, and remove it. Our tokens are invalidated immediately and data collection stops. You can also request full deletion of stored results via the contact form.

No. Report rows are fetched, analyzed in memory, and discarded. We store only the conclusions: findings, evidence summaries, impact estimates, the written report, alerts, and verification deltas. Details are in the Privacy Policy.

Anyone with the report link. Reports live at unlisted, unguessable URLs so you can share them with teammates or investors — treat a link like any confidential document link.

By deterministic code, not AI. Each heuristic compares your measured metrics against versioned industry benchmarks (cited at the bottom of every report) and computes a conservative range. The AI only writes the narrative prose from those already-computed numbers — it is not allowed to invent figures.

No — and be suspicious of anyone who guarantees ad revenue. Ranges assume current traffic holds; seasonality, auction dynamics, and store featuring add noise. Treat them as direction and rough magnitude. Verified before/after measurements are observational, not proof of causation.

Tell us via the contact form — genuinely. Every disputed finding either reveals a threshold we should calibrate or a nuance of your setup the report should acknowledge. This feedback loop is how the agent gets smarter.

It snapshots your account's previous 7 days as a baseline. Seven days later the daily watch measures the after-period and stamps the finding with the verified eCPM and revenue delta — honest numbers about whether the fix actually moved anything.

Each audit is a fresh snapshot of a moving market. eCPMs, fill, and demand shift constantly; a finding shrinking or growing between audits is normal. Old report links stay frozen as generated so shared reports never change under someone's feet.

Each country × format segment's recent eCPM against its own 21-day baseline. A drop of 25% or more opens an alert with the numbers and the likely causes to check, shown in the Watchtower section on your home page. It also runs fix verifications that have come due.

AdMob tells you a metric moved. Florin's alerts are segment-level (US · Rewarded, not “revenue”), compare against your own baseline, list the likely causes in priority order, and connect to fix orders and verification. The goal is one day to diagnosis instead of one billing cycle.

No. It uses the same revocable read-only OAuth grant as audits. Revoking access stops the watch immediately.

Yes, via CSV export. Export a daily historical report with Date, Country, App, Operating system, and Inventory format dimensions plus the standard totals (revenue, impressions, ad requests, unmatched requests, clicks), and a Demand channel report. The import runs the same audit engine; geo and platform checks unlock when those columns are present.

On the roadmap as the first expansion — MAX's APIs support both reporting and (unlike AdMob) real write access, which is where one-click fix application will land first.

A versioned benchmark set (currently calibrated from Appodeal's published quarterly data and industry rules of thumb), cited with version and as-of date in every report footer, refreshed quarterly. As audited accounts accumulate, benchmarks shift to consented, anonymized cross-account percentiles — which are stronger than any public average.

The audit and its shareable report are free. Paid offerings (the daily watch subscription, and recovery-based engagements for larger accounts) are being finalized — pricing will always be shown before you pay anything.

Never of your revenue. For larger accounts we may offer an optional recovery-share deal — a percentage of the verified uplift after fixes, for a limited period, with a cap — as an alternative to the flat subscription. You choose the model; the flat option always exists.

The plan is a free trial period long enough to catch at least one real anomaly cycle. Ask via the contact form and we'll set you up as an early user.

During early access the Google app runs in testing mode, so only allow-listed accounts can sign in. Contact us with your Google email and we'll add you. The “unverified app” interstitial is expected in this phase — click Advanced → Continue.

The connected AdMob account has no reporting activity in the last 30 days. The connection itself worked — come back once an app has roughly two weeks of live traffic.

Something failed mid-run — most often an expired Google session. Sign in again and re-run; if it persists, send us the report link via the contact form and we'll look at the logs.