CHART · CASE NOTE · The self-inflicted plateau

You're paying to reach the patients you already have.

Revenue's been flat for two years and nobody can say why. The spend went up. The "conversions" went up. Here's the mechanism nobody's dashboard was built to show: without proper attribution, the ad algorithm quietly learns to spend your budget on people who were already walking in the door.

T — 1

The mechanism — how the algorithm optimizes you into a plateau

An ad algorithm is a heat-seeking machine. Point it at "conversions" without telling it which conversions are new revenue, and it does exactly what you asked: it finds the cheapest, most certain conversions available. Nothing converts more cheaply or certainly than a person who already decided to call you and is Googling your name to find the number.

Illustrative. Fig. 1 is a mechanism diagram, not a measurement. It carries no data — the "$0 incremental revenue" box is the definition of the problem, not a figure read anywhere: a patient who was already walking in produces no incremental revenue by arithmetic.

T — 2

Two years, two charts, one missing measurement

Same practice, same market, same spend growth. The only variable that changed is whether the algorithm was told which conversions were new patients. On the left: the trap running unchecked. On the right: attribution installed, and the budget redirected at people who don't know your name yet.

✕ Before — the branded trap

Illustrative — not a client's numbers

Diagnosis: the budget bought patients who were already yours. Growth spend and retention spend were the same dollar — and it never left the building.

✓ After — attribution installed

Illustrative — not a client's numbers

Correction: offline conversions taught the algorithm that branded clicks were existing patients. It stopped bidding to reach them and started hunting net-new. The budget never grew — the revenue did.

Illustrative. Both charts are a composite drawn from patterns we see on intake, not a client's numbers. The 40% figure and the shape of both lines stand in for the direction of the mechanism; they are not a measured result, not an industry average, and no engagement is described here. Where this site publishes measured figures rather than illustrations, every one is read in its primary source and carries that source in the row — the reference ranges at /benchmarks, and the register, including which sources we refuse to use and why, at /sources.

The fix is not "stop running branded search." It's giving the algorithm the one fact it was missing: which conversions were new.

T — 3

The correction — three moves that break the loop

None of this requires a bigger budget or a new agency mythology. It requires the measurement layer to tell the truth about which clicks became new revenue, and let the platform re-aim itself.

Attending

Admen '26

Request an examination →