Mobile and App Advertising

Mobile Web Versus In-App

Mobile web advertising follows standard display and video mechanics with mobile-specific constraints — smaller viewports, variable connection quality, and higher sensitivity to load time.

In-app advertising is a separate inventory pool with its own formats: rewarded video where users opt in for an in-game benefit, interstitials between activities, and native placements within feeds.

Rewarded video deserves attention. Completion rates are high because the exchange is explicit, and users report it as the least intrusive ad format.

App Install Campaigns

Google App Campaigns and Meta app install campaigns automate placement and bidding across their networks. You supply assets and a target cost per install or per in-app action.

Bidding to install optimises for volume and frequently acquires users who never open the app again. Bidding to an in-app event — registration, first purchase, day-seven retention — costs more per acquisition and produces materially better cohorts.

Optimise to the deepest event that has enough volume to train the algorithm.

Mobile Attribution

App attribution does not work like web attribution. Mobile measurement partners such as AppsFlyer and Adjust sit between ad networks and the app to resolve which click or impression produced an install.

Apple's App Tracking Transparency changed this substantially on iOS. Without user permission, deterministic attribution is unavailable and SKAdNetwork provides aggregated, delayed and limited data instead. Android retains more granular measurement, though Privacy Sandbox is moving in a similar direction.

Fraud in Mobile Advertising

Install fraud is significant and takes recognisable forms: click injection, click spamming to claim credit for organic installs, SDK spoofing, and device farms.

Signals worth monitoring: implausibly short click-to-install times, abnormal concentration of installs from few devices or IPs, and installs that never generate any in-app activity.

Use a measurement partner with fraud detection, and treat any source with excellent install volume and no downstream engagement as suspect.

Attribution After the Identifier Changes

Mobile attribution was built on a stable per-device advertising identifier. That assumption no longer holds, and the practical consequences are larger than the policy summaries suggest.

On iOS, access to the identifier requires user permission through the app tracking prompt, and consent rates are well below universal. On Android the identifier can be reset or withheld. So a material share of installs cannot be deterministically matched to a click.

What replaces it is aggregated, delayed and privacy-constrained: postback frameworks that report conversions in coarse buckets after a delay, with thresholds below which nothing is reported at all. The operational consequences are what matter — small campaigns can fall below reporting thresholds and appear to produce nothing; optimisation windows lengthen, so fast iteration on creative becomes unreliable; and post-install event data is limited, which weakens the value-based bidding that mobile campaigns depend on.

Plan for fewer, larger, longer campaigns rather than many small fast ones. The measurement environment now rewards that shape.

Mobile Fraud, and What Actually Reduces It

App install advertising carries a distinctive fraud problem because the payable event — an install — is cheap to fabricate.

The recognisable forms. Click injection, where an app on the device detects an install beginning and fires a click moments before completion to claim attribution. Click flooding, sending enormous volumes of clicks so that some organic installs attribute by chance. Install farms, real devices installing at scale. SDK spoofing, fabricating the install signal without an install happening at all.

The signals that expose them are distributional. Click-to-install time clustered near zero indicates injection; a very long flat tail indicates flooding. Conversion rate far above the network norm is a finding, not a success. Post-install engagement near zero is the clearest signal of all — fraudulent installs do not open the app twice.

What reduces exposure: pay on a post-install event rather than the install itself, which removes most of the economic incentive at a stroke; use a measurement partner's fraud filtering and understand what it does and does not catch; agree rejection terms in the insertion order before spending; and cap and monitor by sub-publisher, since fraud concentrates in a small number of sources that a network-level view averages away.

Sources

What each claim on this page rests on. Entries are typed so you can see which are primary.