Measurement

Two systems measuring genuinely overlapping things with different definitions. As with the other centre's own paid billboard, the answer is to pick one system of record per question, rather than trying to reconcile them.

What the advertising tool reports

Installs and in-shop conversions credited to the advertisement, with cost, and performance against whichever target you set.

The centre's own diagnostic for a constrained order: your achieved cost closely tracking or occasionally exceeding your target, or your target return consistently higher than what you're actually achieving.

Reporting at the level of individual material exists, but it's deliberately coarse, the system itself is deciding how to allocate attention, and the reporting reflects that rather than offering you real per-item control.

What your own back-office reports

Traffic split into this centre's own search, its own browsing, ads and referrals, plus people who visited your listing and first-time installers.

A specially-targeted second window's own performance only shows up once it's crossed a real minimum amount of visitor volume, and the centre notes that these targeted windows typically pull in fewer visitors than a broader campaign would.

The main measurement here also shifted recently from actual acquisitions to clicks. Comparing any campaign result from before that shift to one measured after it compares two genuinely different events.

Why the two numbers will never actually match

Genuinely different attribution models and windows. The advertising tool credits things on its own model; your own back-office counts listing visits and installs on its own separate definitions entirely.

Your own back-office counts people; the advertising tool counts conversions.

The "ads and referrals" bucket isn't just your own campaign. It contains every kind of paid traffic and referral, including entirely different networks.

The advertising tool sees conversions your own listing simply doesn't. In-shop actions after the install, which your listing's own acquisition reporting never covers at all.

Don't try to build a reconciliation between them. Use the advertising tool for optimising the campaign itself, your own back-office for the shopfront-side conversion of windows you actually control, and an outside measurement partner for comparing across every network at once. Always say which system any given number actually came from.

What's genuinely measurable from the being-found-and-chosen side

Conversion on a dedicated, specially-targeted second window. Traffic arriving there is genuinely separable, so the shopfront-side conversion for that specific advertisement is real, subject to the usual minimum volume threshold.

Whether paid is genuinely displacing your unpaid results. Your own back-office's traffic split lets you watch your own unpaid search and browsing numbers while ad spending changes. The exact same reasoning as on the locked side's own paid billboard, and the only genuinely honest answer is still a real holdout.

The install referrer. This centre's own mechanism for passing campaign information through to the shop once it's actually installed.

What tends to go wrong

Trying to reconcile the advertising tool and your own back-office. Genuinely different definitions underneath.

Reading the "ads and referrals" bucket as if it were just your own campaign. It's every paid source at once.

Comparing across the point where the main measurement changed.

Expecting real per-item credit. The system allocates on its own; the reporting simply reflects that.

Judging a specially-targeted window purely on raw volume. The centre states plainly that these generate fewer visitors, and may sit below the reporting threshold entirely.