View-through attribution

What view-through attribution measures

Some ads are not built to be clicked. A person watches a video, closes it and installs the app later from the store. Click-through attribution records nothing for that ad. View-through attribution records the impression as a touchpoint and lets it take the credit if nothing stronger exists.

The ad network reports the impression to the MMP, usually server to server, because there is no tap on a tracking link to carry it. The MMP then holds the impression for the length of the view window and waits for a first open from the same device.

How a view competes with a click

A view normally ranks below a click: a qualifying click wins even when a view is more recent, and a view can win only when no click is inside its window. The full order of checks is in how last-touch attribution picks a winner.

What a view needs before it can be matched

A click leaves a trail: a link was opened, and it can carry a click ID. A view opens no link, so it usually has less to match on. How a view is matched depends on the channel and the platform. Ask your provider which channels it can match views for, and by which method.

The risk: paying for installs you already had

An impression is cheap to serve and easy to serve widely. A network that shows an ad to a very large audience will, by chance alone, show it to people who were about to install anyway. With a long view window, each of those installs moves from the organic column to the paid column, and the channel's cost per install looks better than it is.

Where the platform sets the rule

You do not control view-through everywhere. Apple's privacy frameworks apply their own logic and report the result to you. One documented example: in AdAttributionKit, the system does not create re-engagement postbacks from view-through ads.^[1] The guide to AdAttributionKit covers that framework. Your own view window applies only to installs the MMP matches itself.

How to set it up

1. Leave view-through off for channels where people normally click, such as search.
2. Turn it on per channel for video and display, with a window much shorter than your click window. One day is a reasonable place to start testing.
3. Compare organic installs for the two weeks before and after. If organic falls by about what the channel gained, shorten the window.
4. Report click-through and view-through installs in separate columns, so nobody reads them as one number.

Airbridge's plan comparison lists a lookback window that you configure per channel and match type, down to hours or minutes.^[2] It does not say whether a view window is one of those settings, so confirm that before you rely on it. Next, read how self-attributing networks report, because some networks keep their view data and attribute for themselves.