80% of an app campaign's success is decided before you create it. This is the list I would go through with a client before authorising the first euro of spend. If any of the first four fails, you don't launch: you fix it first.
1. Measurement inside the app (blocking)
Is the SDK installed and tested — Firebase or an attribution provider (AppsFlyer, Adjust, Singular, Branch)? Do events arrive in the console in real time? Without this, the campaign is blind: it will count installs and little else.
2. The events that actually matter (blocking)
An install is not a business goal; it is the beginning. Define and measure the in-app events that mark value: sign-up, tutorial completed, first purchase, subscription, revenue. It is exactly the signal ladder from Lead Scoring applied to an app: the install is the "raw lead" and the purchase is the sale.
3. Linked accounts (blocking)
Google Ads with Google Play (and with Firebase or your attribution provider) and, on iOS, whatever your setup requires. Without linking, conversions don't travel: you will see clicks and installs without knowing where they came from or what they did next.
4. A presentable store listing (blocking)
Icon, title, first screenshots and video. Remember the previous lesson: this is your landing page and its conversion rate multiplies or divides your cost per install. Sending traffic to a neglected listing is paying to show your worst face.
5. Enough creative inventory
This is where almost everyone falls short. Public guidance from Google and specialist agencies points to several text assets, a good handful of images in different aspect ratios and several videos (portrait, landscape and square), because each surface — Search, Play, YouTube, Display, Discover — asks for a different format. With two images and no video, your campaign cannot even show up in half the places it could.
6. A realistic budget for your chosen bid strategy
App campaigns need volume to learn. As a widely quoted public reference: a daily budget of around 50 times your target cost per install for install campaigns, and on the order of 10 times your target CPA when optimising for in-app actions. If your budget doesn't reach, don't build five campaigns: build one.
7. An agreed learning expectation
There will be one to two weeks of volatile results while the system learns, and every large change restarts that countdown. This is agreed with the client or your boss BEFORE, not on the day the CPI spikes.
8. A plan for reading results
Decided in advance: which metric rules (cost per install, cost per in-app event, return), in which window it is judged, and how often you touch anything. Without this plan, the campaign gets optimised on impulse.
💡 Ninja trick: if your app is new and has no volume to optimise for purchases, start by measuring a frequent intermediate event (sign-up, first level completed, first long session) and climb the ladder once there is data. Same principle as Lead Scoring: feed the algorithm signals that happen early and often.
⚠️ Pitfall: launching "as a test" with €10/day and no in-app events. It tests nothing: there is no volume for the system to learn and no data to judge it. It is money that buys noise — and it generates the false conclusion that "Google Ads doesn't work for apps".
What you should remember
- Four blockers: SDK, in-app events, linked accounts and a decent listing. Without them, you don't launch.
- Varied creatives from day one: they are the system's ammunition.
- A budget sized to learn, not to "test".
- A 1-2 week learning period agreed, and a written plan for reading results.
At the Advanced level you have the full module: how they work inside, measurement and iOS, structure, the bidding ladder, creatives, ASO, re-engagement, scaling and install fraud.