US$54,718 of dental Meta spend, and the metric that actually predicted cost

Campaign A bought clicks at $1.36 and leads at $41.63. Campaign B bought clicks at $2.18, 60% more per click, and leads at $33.59, 19% cheaper. Anyone optimising toward cost per click alone would have kept funding Campaign A and starved the campaign that was actually working.

Frequency sat at 3.75 for Campaign A against 2.54 for Campaign B. The gap is worth watching, not treated as the cause: these are two line items in one account over one period, and frequency moving alongside a cost difference is not the same as frequency producing it.

Three signals were tested against cost per lead across every line item above $500 in spend. Leads per click correlated at -0.569, by far the strongest relationship in the account. Frequency correlated at -0.384. CPM, the number most often quoted back in a monthly report as if it stood for performance, correlated at +0.098: effectively no relationship at all.

That is the one worth sitting with. An account can be reported entirely in CPM and still tell you almost nothing about what a lead is going to cost. This is one account over one date range, and only the line items above $500 in spend within it: a signal worth acting on, not a law. It describes what happened here, not what will happen everywhere.

The best line item in the account bought leads at $9.81. The worst bought them at $278.79, on almost the same spend, a 28-fold spread inside a single account. The best line item ran a leads-per-click rate of 23.3% and a frequency of 1.40.

The account-wide average of $39.08 describes neither end of that range. It is an average of a $9.81 line item and a $278.79 line item sitting side by side, and reporting only the average hides the decision that actually mattered: which line items to keep funding and which to cut.

1,400 leads became 1,092 qualified leads, a 78% qualification rate, and 1,092 qualified leads became 470 new patients, a 43% close rate. Qualification is the clinic's judgement, not the platform's: it is the clinic deciding a lead is the right service, contactable, and not already a patient. The close rate that follows is the clinic's front desk, not the ad account.

Cost per new patient came out at $116.53. That figure is calculated on unrounded counts, so dividing the published spend of $54,718.34 by the published patient count of 470 lands a few cents away from it. Both are correct: they are two different divisions of the same underlying numbers.

What transfers from this account is a method, not a price. Watch leads per click rather than CTR or CPM: it was the metric that predicted cost per lead here, and CPM did not. Split the account at whatever leads-per-click line separates the top half from the bottom half, and cut the tail rather than letting it coast on its historical spend.

Expect the spread inside an account to be wide, not narrow: this account's best and worst line items landed on very different prices on comparable budget. There is no promise or projected figure to draw from this page, and no New Zealand number either: auction costs move with the market, the season and the offer, and none of those transfer.