How it works
Most ad accounts bleed money where nobody is looking. We look.
Drop a Meta CSV, read the verdict, keep the dashboard, get a weekly diff. A watchtower on a connected account is what comes after — read-only, flat-rate, and built so it can't ever touch your account.
Step 1
Your CSV
Exported from Meta Ads Manager.
Step 2
$4,200Your verdict
Ranked findings, math beside each.
Step 3
14%Your dashboard
History + spend efficiency. Saved.
Step 4
Weekly diff
Email only when something changed.
Q1 2027
Watchtower
Continuous, read-only.
The free path
Drop a CSV. Read the verdict.
A 60-second CSV export on your side. The engine runs all 24 rules against every campaign, ad set, ad, audience, and placement in your account. No sampling.
- 01
Export your Meta CSV.
In Meta Ads Manager: Ads reporting → Export → CSV. Pull 30+ days with the day breakdown on. No app review. No OAuth. No data leaves Meta until you choose to upload.
- 02
Drop it on us — or send it to Claude.
Upload at /audit, or call
run_audit_from_csvfrom Claude Desktop, Cursor, ChatGPT, or any MCP-aware client. The MCP server is free to install. Same engine, your choice of surface. - 03
You get a verdict.
Ranked findings, dollar amount at risk on each, a confidence number on each, plain-English fix steps, and the math behind every number. The methodology page is public for every rule that fires.
48,720
data points examined per audit on a 60-day, $30k/mo account
47
statistical signals computed — frequency trends, CPA ratios, overlap matrices, dominance scores
287
individual checks run — all 24 rules against every eligible entity
What we look at that most tools miss
Per-ad conversion rate against siblings, not the account average.
Catches winners that are only winning because Meta over-served them. Comparing to the account mean hides the false winners.
An audience overlap matrix across every pair of custom audiences.
For an account with 12 audiences, that's 66 pairwise checks. Most tools do this at the campaign level, which is too coarse to catch it.
Per-placement CPA inside every ad set.
Aggregated up to the campaign level, placement waste hides. We look ad-set by ad-set.
Short rolling windows, not lifetime averages.
A lifetime CTR hides week-over-week saturation. We compute the deltas, so a slow fade still gets caught.
Built for the marketing manager defending a tool to a CEO, the growth lead who's done with false positives, and the agency owner walking into a client review. Read the full methodology, every rule →
Sign up to keep it
Your account turns the audit into a habit.
The audit is one-shot. The account is what compounds. Every audit saved, comparable month-over-month, with one number on the homepage that tells you whether you're running cleaner or dirtier than last time.
History + comparison.
Every audit you run is saved. Month-over-month deltas are computed by the engine, not eyeballed.
Your spend efficiency, charted.
The same 14% benchmark we use everywhere — but on your accounts. One number, with the math one click away.
Weekly “what changed” email.
Re-audited on your CSV cadence. We email only when a finding fires, resolves, or moves materially.
Free with any account · no card.
Good morning · today
Welcome, there.
Across your accounts, $4,200/mo is currently on the table. The biggest single fix is below.
Spend efficiency
Fixing these is equivalent to running with 1.16× efficiency at the same budget.
Next best action
Meta is over-serving one ad — and it's not your best one.
rule.false_winner_meta_bias
$1,840/mo
90% confidence
Engine ready · 24 rules · re-run any time
Q1 2027 · planning
Between every audit, the watchtower watches.
Connect your Meta account, read-only. The watchtower does everything an attentive consultant would — except it never sleeps, and it never forgets.
Capability
Continuous monitoring.
All 24 rules re-run every six hours against your live account — and within minutes of any meaningful change (new creative, budget shift, fresh campaign). You won't have to schedule audits. The engine is simply on.
Capability
Alerts where you actually look.
Email, Slack, WhatsApp, Telegram. Routing by severity: critical to WhatsApp, weekly digest to email. Every alert links back to the row of data that triggered it.
Capability
Account memory.
The watchtower remembers what you've fixed. When the same problem starts re-emerging on a re-launched retargeting set, it tells you it's back — not that it's new.
Capability
Read-only, hardcoded.
We ask Meta for the ads_read scope only — never write. There is no “apply optimization” button anywhere. We surface findings; you make the call.
We'll meet you on the channel you actually check
Weekly digests + the audit trail.
Slack
Shared with your team or agency.
Critical findings, before tomorrow morning.
Telegram
Founders who live there.
The 24 rules are public. The watchtower is what we'll actually sell — because operating those rules continuously, recalibrating against real accounts, and routing findings to the channel where you'll act on them is the part nobody replicates by re-reading the methodology page.
The boundaries
Three things we won't ever do.
01
Never write to your account.
Read-only, hardcoded in the source. No “apply optimization” button. You see the finding, you make the call.
02
Never charge on your ad spend.
Flat-rate, always — never a percentage of spend. Whether you spend a little or a lot, the math costs the same to run.
03
Never review your creative.
Use Motion for that. We stay on the parts of Meta Ads where the math decides what's wrong. Creative is a taste call, not a math call.
Try it
See it on your account.
The first audit is free. The account makes it a habit. The watchtower comes next.
Want it inside Claude Desktop, Cursor, or ChatGPT? Install the MCP server → Free for CSV audits; subscribing unlocks live watchtower tools through the same install.