Two weeks debugging the wrong thing

A short account of how long it takes to find a failure that nothing reports, written so you can skip that part.

6 August 2026 · PAD team

Day one to three: the creative

Client said the ads were fine and nobody installed. CTR was normal, spend was pacing, installs flat. So we did what everyone does and blamed the hook, then the thumbnail, then the first three seconds.

New variants performed identically. That should have been the first clue: when everything you change makes no difference, you are changing the wrong thing.

Day four to six: the store page

Screenshots, subtitle, first impression. Reasonable suspects, since the funnel ends there. We could not test the hypothesis properly because store page conversion looked fine for other traffic sources.

Day seven to nine: attribution

Convinced something was misconfigured, we went through the tracker setup twice. Everything was correct. The tracker was faithfully recording clicks that we now know went nowhere.

Day ten: the accident

Someone tapped the ad from their own phone, inside Instagram, rather than pasting the URL into Safari like we had been doing all week. White page. Nothing loaded.

That was the whole investigation, and it took ten seconds once anyone did it the right way.

The test that would have saved thirteen days

Tap your own live ad from inside the app on a real phone, before changing anything else. Not from Safari, not from the simulator, not from a desktop preview.

Every hour we spent was on the assumption that people were arriving and not converting. Nobody checked whether they were arriving.

What we do first now

On any new account with weak install numbers: phone test, objective comparison, platform split. Twenty minutes, before touching a single creative.