Rethinking cost per started case rather than cost per lead

Author : Reputation Elevation | Published On : 16 Oct 2026

The quarterly budget meeting had reached the part where two campaigns were being defended. One had produced enquiries at nineteen dollars and was described as the account's best performer. The other produced them at eighty one and had been on the chopping block since June. When somebody traced which enquiries had become children in treatment, the ranking reversed completely.

Adopting a better metric is the easy part. What follows is harder, because the number immediately starts contradicting decisions that have already been made and defended in public.

The reallocation that follows

The first casualty is usually broad, cheap traffic. Campaigns built around general autism information produce volume at low cost and start cases rarely, because a reader looking for a definition is not a parent choosing a provider. They are not worthless, they simply belong in a different part of the plan with a different expectation attached.

The same reversal catches early-stage queries. Searches about development and milestones are legitimate and large in volume, and somebody checking the CDC's developmental milestones for a sixteen month old is typically a year away from choosing a provider, if they ever become one at all. Buying those clicks with the same bid as a query naming a plan and a town guarantees a flattering cost per lead and a poor one per case.

Branded terms usually move the other way. They look expensive on a per-lead basis because the volume is small, and they convert to started cases at rates nothing else approaches, because the person typing a practice name has usually already been recommended by a paediatrician or a neighbour.

Bidding towards an event the platform cannot see

The complication is structural. The started case happens weeks later, inside a clinical system, and cannot be sent back to an advertising platform in any form that would identify a person. So the account cannot optimise directly towards it. Practice owners running the same calculation on paid social can find out more about Reputation Elevation and how Meta campaigns are handled for ABA practices.

Two workable approaches exist. The first is offline conversion import keyed to a non-identifying reference, which gives the platform a delayed, aggregate signal that this click produced value without telling it anything about who. The second is proxy conversions, chosen deliberately: a completed benefit verification rather than a form fill, a booked assessment rather than a phone call of any length.

Both demand that the practice decide what it is willing to pay for a started case before the bidding is set, which is a finance decision wearing a marketing label.

The learning period deserves respect here. Optimising towards a rarer event with a long delay needs more volume and more patience than optimising towards form fills, and a practice that switches and then panics at week three will have spent the learning cost twice.

The arguments it settles and the ones it starts

It settles the argument about channel quality, usually within one quarter, because the figures stop being debatable. It also settles the argument about whether to keep paying for brand terms, which otherwise recurs annually and is never resolved by anyone's opinion.

What it starts is less comfortable. Once a started case has a price, intake delays have a price, authorisation friction has a price and clinical capacity shortfalls have a price, and those costs sit with departments that never previously appeared in a marketing conversation. The metric does not respect the organisational chart.

That discomfort is the signal it is working. A number that only ever criticises the advertising is not measuring the business, and the practice that put nineteen and eighty one in the right order was the one that stopped treating marketing and operations as separate accounts.