A note on retention.

I spent the morning trying to make sense of a CPA spike that looked like a problem at first glance. For three days, it pulled me into dashboards, reports, attribution windows, and campaign notes. I almost rewrote a perfectly good campaign because the numbers made it feel like something was wrong.

The spike showed up on Monday. Cost per acquisition jumped across multiple channels, but nothing obvious had changed. No new campaign structure. No creative swaps. No landing page updates. Just a number moving in the wrong direction with no clear reason attached.

So I did what you do when the data starts acting strange. I pulled the raw reports, checked the platform dashboards, compared attribution windows, and looked for anything that could explain the shift.

Everything looked normal until it didn't.

The issue was not the campaign. It was the way the conversions were being counted. When the data changes shape, ask whether the campaign moved or the ruler did.

By Wednesday, I was ready to move forward with a new creative test. The team had already prepared a few variations, and we were close to launching.

Then I caught it.

The spike was not only showing up in our primary attribution model. It was also appearing in our backup view-through window, which had quietly been extended from 7 days to 14 days.

That changed everything.

The CPA was not actually getting worse. We were just counting more conversions as paid that would have been considered organic under the old window. The campaign was still healthy. The measurement had shifted underneath it.

So I paused the creative test, kept the campaign running, and wrote a note to the team explaining what changed so we would not chase the wrong fix again.

This was one of those reminders that retention is not just about keeping users. Sometimes it is about keeping your patience long enough to understand what the numbers are really saying.