Conversion capture rate

You meet it when the sales team and the ad platform describe the same month completely differently. Sales say the pipeline is healthy and the leads keep coming. The ad account says each one cost several times what the business can afford. Both are reading their own records honestly.

It is most common on accounts that generate leads rather than online sales: a form, a phone call or a chat that becomes a customer days or weeks later, somewhere the platform cannot observe. It is nearly universal on accounts whose offline conversion import has silently stopped working, which the interface does not announce.

Capture rate is the share of real outcomes the platform manages to record. Anything that stops a lead being sent back to the platform lowers it: a broken or expired offline import, a consent banner blocking the tag, a form on a subdomain the pixel does not cover, a phone call taken outside a tracked number, a CRM whose lead source field was filled in by hand.

The cost inflation follows mechanically. Spend divided by a fifth of the outcomes is five times spend divided by all of them. This is not an estimate or a modelling assumption. It is division, and it means capture rate and reported cost per acquisition are the same fact expressed two ways.

The reason it goes unnoticed is that the platform never signals a missing conversion. An account importing nothing looks exactly like an account whose campaigns stopped working, and the interface presents both as a performance number.

Automated bidding optimises towards the conversions it can see. On a low capture rate it is not merely reporting a wrong number, it is actively learning from a biased sample: whichever segments happen to convert in ways the platform can observe get bid up, and whichever convert in ways it cannot get starved, regardless of which are worth more.

On top of that sits the human decision. A target set from an inflated cost per acquisition throttles the account. Campaigns get paused, budgets get cut, and the channel gets written off, all against a figure that was several times too high before anyone looked at it.

Compute the capture rate before you read any cost number from the interface. Count the leads your own records attribute to the platform over a settled period, divide by what the platform counted, and you have both the health of the measurement and the correction factor in one number.

Then fix the import rather than the bid. Restoring the feed is usually a day of work and it repairs the reporting, the bidding signal and the historical comparison at once. Adjusting targets to compensate for a broken import fixes none of those and hides the fault.

There is a related trap worth naming while you are in there. Campaign-level conversion goals mean an account-level primary goal flag no longer describes what a campaign optimises for. Check the campaign, not the account, before concluding what it is bidding towards.

It is not the same as over-counting. Capture rate describes outcomes the platform missed. A platform can be missing genuine leads and simultaneously over-claiming credit for the ones it does see, and most lead-generation accounts have both.

It is not a reason to distrust the platform's relative reporting. Even at a low capture rate, the comparison between two keywords inside one account is usually still directionally useful, because the same blindness applies to both.

It is not a fixed correction. The rate moves whenever consent rates, form placement, sales process or the import itself changes, so it is measured on a schedule, not once.