What are fake in-app events?
Fake in-app events are fabricated actions, such as sign-ups, trials or purchases, reported to an attribution platform without a real user performing them. They inflate post-install metrics and the payouts tied to them.
How they are created
- SDK spoofing: requests are sent directly to the attribution server, imitating a real SDK. See SDK spoofing.
- Bots: scripted traffic installs the app and completes simple actions, often through a device farm or emulator network.
- Manipulated apps: a modified app build fires events that a real user never triggered.
Why it matters
- CPA payouts: where a partner is paid on cost per action, a fake event is a fake invoice line.
- Misleading quality: fake traffic appears to convert, so a poor source can look like your best one.
- Skewed optimisation: ad networks learn from the events they receive. Fake events can teach their bidding to chase similar fake traffic.
Signals to watch
- Events that fire in bursts, or at identical intervals after install.
- A high event rate paired with poor retention.
- Identical event sequences, values or parameters across many users.
- Revenue events with no matching payment on your server.
- Events arriving with no preceding session.
Who is most exposed
- Apps that pay partners on registrations, trials or first purchases.
- Campaigns where the optimisation event is cheap to trigger, such as a sign-up with no verification.
- Sources with sudden volume spikes and no history.
Practical defences
- Send only events that matter, and keep the list short. See in-app events.
- Validate revenue and sign-up events against backend records.
- Compare event rates with retention for each source before paying on them.
- Apply fraud rules to the underlying install, and consider holding back signals for flagged installs. See rejection postback.
How Linkrunner relates
Linkrunner's Fraud Control includes SDK spoofing and SDK spoofing with IP mismatch rule templates, with a Tag only or Tag and withhold postbacks response on each. Fraud Control is included on every plan, and attribution is never changed.