Bidding and signals

You don't bid on anything specific here at all. You state a goal and a target, and the centre's own system handles everything else.

The goals you can actually set

A target cost per install. For an order purely aimed at getting new people to install the shop.

A target cost per action. For an order optimising toward something someone does inside the shop, rather than the install itself.

A target return on spend. For an order optimising toward actual revenue. This can only genuinely be assessed if the value of that action is actually known, so whatever you're optimising toward has to be passing real value back from your own analytics.

That last point is the common failure: an order aimed at return on spend, with no real value ever passed back, is optimising toward a number it literally cannot see.

The budget rule

For an install-focused order using a target cost per install, your daily budget should be at least around fifty times that target.

This is the single most useful number in this whole section. A modest target cost per install needs a genuinely large daily budget behind it. Below that, the order simply can't gather enough actual results to learn from, and it will underperform in a way that looks like a targeting problem but is really just a volume problem.

It also means this kind of advertisement has a genuine minimum viable spend. A small budget spread thinly across several orders performs worse than one order properly funded. And that's a real, legitimate reason a team might be better served by the simpler pay-per-install billboard on the other centre than by an underfunded order here.

Diagnosing a constrained order

The centre's own diagnostic: if your actual achieved cost is closely tracking or occasionally exceeding your stated target, or if your target return is consistently higher than what you're actually achieving, the order is constrained by its own target.

Being constrained like this means your target is simply too tight for the real auction, so the system can't buy any real volume without breaching it. The fix is raising the target, or accepting less volume. Not adding more material, and not restructuring anything.

The signals you actually control

There's no list of words here, so your real influence comes from elsewhere entirely:

What counts as a real conversion. What you tell the system to actually optimise toward. An order optimised purely toward installs and one optimised toward an actual purchase buy you two genuinely different kinds of visitor. This is the single highest-leverage decision in this whole area.

The value of a conversion. Required for a return-on-spend goal, and it genuinely has to come back from your own analytics.

Your actual raw material. The creative signal covered on its own page.

Your own shop listing. This kind of advertisement sends traffic straight to your regular listing, so everything that governs how well that listing converts applies here too. And with a specially-targeted listing, the landing page itself becomes a real choice you make.

Your targets and your budget. The goal, and the real constraint around it.

What tends to go wrong

Underfunding relative to your own target. Your budget should be at least around fifty times your target.

A return-on-spend goal with no real value tracking. Optimising toward a number the system can't actually see.

Treating a target-constrained order as a creative problem. The centre names this exact diagnostic itself.

Splitting a small budget thinly across many orders. Each one learns almost nothing.

Optimising purely toward installs and then expecting genuinely high-quality visitors. You get exactly what you asked for.