How I put a price on a leak

The most useful thing I do is not finding problems — any checklist finds problems. It is pricing them, because a founder with eleven findings and no prices cannot prioritise, while a founder with three findings worth $520, $310 and $90 a month knows exactly what to approve first. Here is the arithmetic, shown in full so you can check my work.

The general form

Every leak price is the same shape: (spend exposed to the problem) × (the efficiency lost because of it), measured against the account’s own baseline — never an industry average. The baseline is what makes the number honest: the question is always "what did this money earn, versus what money like it earns elsewhere in this account?"

Worked example: creative fatigue

A creative spent $2,000 last month. Its click-through over its first three weeks averaged 1.8%; over the last two weeks it has run at 1.1% with CPMs steady. The decayed fraction is roughly (1.8 − 1.1) ÷ 1.8 ≈ 39% of its former efficiency, applied to whatever it spends now. If it keeps spending $1,500 a month in this state, the leak is about $580 a month against its own demonstrated capability. Not a guess — its history is the proof of what the audience used to do.

Worked example: audience overlap

Two ad sets target audiences where 40% of the people are the same. The cost is not 40% of spend — that is the exaggeration version. You are not bidding against yourself either: Meta enters only one of your ad sets in each shared auction. The real cost is the delivery the other ad set loses, the extra time it can spend in learning, and the duplicated frequency. That prices out well below the headline overlap — but it is real, recurring, and fixable in an afternoon with exclusions or by combining the ad sets.

Why I round down

• Every estimate uses the conservative end of its range — an audit that overpromises gets one approval, then loses trust forever.
• Fixes interact: pausing a fatigued creative changes the math for the ad sets around it. The next audit re-prices what is left rather than summing a static list.
• Some losses are unknowable: what the budget would have earned in the best alternative placement is an estimate, and I label it as one.

What a priced audit changes

Prioritisation stops being a debate. The $520 leak gets fixed before the $90 one regardless of which is more interesting to argue about, and the approval decision becomes arithmetic: a fix that recovers more in its first month than the tool costs in a year is not a difficult call. That is the entire reason I refuse to ship a finding without a price tag — an unpriced finding is an opinion wearing a clipboard.