App Install Ad

On the web, clicking an ad opens the page. An app doesn't work that way. The person who clicked lands in an app store, waits for a download, then has to find the app and launch it before reaching the first screen. There's an extra threshold sitting in the middle.

Right after Pagedeck shipped its mobile app, Ravi ran the same ads that worked for web signups. Same creative, same copy, different outcome. Plenty of people reached the store listing, but only about three in ten finished the download, and half of those opened the app even once. Two drop-off points existed that the web never had, and running against them called for a format that counts them.

The objective is fixed at one install, and the ad platform steers impressions toward people likely to complete one.

1. The click goes to an app store rather than a browser. This is why creative that shows the actual app screens performs better: when the screens in the ad match the screens on the store listing, fewer people stop in between.
2. Success is counted at first launch, not at download. Somebody who downloaded the app and never opened it has started nothing.
3. The cost measure is cost per install (CPI) — what one install cost. At $2.30 per install, a thousand installs came to $2,300.
4. An install is the beginning, not the goal, so teams usually move the objective one step later. Optimising for completed signup or a first document raises the install price and changes how many people stay. At Pagedeck, 210 of every 1,000 installs produced an account that created at least one document.
5. Measurement is harder than on the web. The ad click and the first app launch are recorded in different places and have to be joined, and device identification limits degrade the accuracy. How far after the click still counts is settled by the attribution window.

"Installs are $2.30, but per account that creates a document it's $11."

The cost to an install and the cost to someone actually using the product are different numbers.

"Put the app screens in the creative — it has to match the store listing."

Reducing the number of people who stop between the ad and the download.

"Skip install ads and go with search ads. This is a product people use on the web."

Without the install threshold, a different format fits.

What Ravi concluded after three months was that the ranking inverts. On install price alone, the cheapest network came first; add the share of installs that went on to create a document and that same network came last. Producing cheap installs and bringing in people who use the product turned out to be different things, and both numbers appeared side by side in every report afterward.