mFilterIt has launched a new algorithm to improve fraud detection and fight against Ad-Fraud. One simple and easy method launched by mFilterIt is using Attribution Type to identify and clean fraud.
We are explaining the rule here on the public forum to encourage advertisers and attribution platforms to start using this in their systems (ideally this check should have been straight forward and embedded into attribution platforms from the start). Our fight against AdFraud only gets strengthened by the industry coming together and working together. This article is in that direction of giving the industry the push it needs to start demanding better technology, attribution and improving the eco-system for everyone to work better.
Attribution in App world is done using three methods (in following sequence) :
- Referral String Based Match: The default and the top priority check involves the flow where user clicks on an Ad, redirects to App Store with a transactionID and when he downloads and opens the app, the App Store returns the same transactionID, resulting in a 100% accurate and deterministic match. Normally, this should be the most widespread mechanism of tracking and contribute majority of the installs
- DeviceID Based Match: The next in priority, requires AdvertisingIDs to be passed in the click to the attribution platform, which then compares with the AdvertisingIDs received at the time of install and open. This is again 100% accurate and deterministic. However, it should only be triggered if Option # 1 is not valid. The ONLY scenario where this should be triggered is, when the user clicks and opens the App Store, but does not download the app at that point, but does it later on (maybe after a few hours). Since the App store was opened later again, the Referral Match no longer works. Hence DeviceID based match. Now (and here is the crux of the check) this should be a much smaller component of the overall matches, and there is no logic for this to work in the first initial few minutes! The probability of a user to click on an Ad, go to App store, but leave it, then open the app store AGAIN in a matter of a minute and download the app, is simply too low.
- Fingerprinting Match : This is the last in priority and only used typically as last resort. Generally it is useful for Web Based matches, where DeviceID may not be available by the publisher to pass in the click.
The Fraud Rule :
Since DeviceID based match can only be on a lower priority than Referral Based match, it CANNOT be explained (apart from fraud) where click-to-conversion time difference is extremely low (e.g 2 mins). Typically, publishers who are trying to capture installs in the background will use this, since the Referral Based solution will not work in the background. Or bots will use it to fire clicks and then trigger installs bypassing the App Store (where again the Referral Based option does not work).
This simple rule is going live in all our clients in the next 1-2 weeks and we recommend all attribution platforms to also start looking at this to improve fraud detection and improve attribution. Currently, with the advent of popular terms like “Machine Learning” and “AI”, simple checks like these can actually improve fraud detection much better instead of focussing on fads.
About mFilterIt :
The industry leading fraud detection solution which has been working in this space for the last 3 years (much before it became a fad for all to talk about!). We are the ONLY neutral, dedicated and independent fraud detection solution, helping advertisers save millions of dollars annually from being wasted. mFilterIt offers advertisers a common platform across app and web platforms, ensuring fraud detection, prevention as well as brand safety! Check us at mfilterit.com