Distribution Modeling

Adjust’s Distribution Modeling proactively protects against click spam. To do this, Adjust calculates the statistical likelihood of a click being the result of click spam, and filters out engagements that don’t meet our standards. This means fraudulent activity never skews your data set and you can confidently identify your valuable organic users.

Growth solution:
The Fraud Prevention Suite is available as an Adjust Growth Solution. To get Fraud Prevention on your account, contact sales@adjust.com.
Tip:
Learn how to set Distribution Modeling for your app.

How it works

At Adjust, we define click spam as illegitimate click activity. For fraudsters, the goal of click spamming is to poach attributions from your organic users - that is, to have a certain number of your organic installs falsely attributed to a fraudulent campaign. This way, these campaigns appear to generate high volumes of valuable users.

Click spam can take various forms. It can range from traffic sources that send views as clicks to servers that send catalogs of artificial clicks. Another common example is when an app invisibly loads and clicks ads in the background.

Adjust’s Distribution Modeling uses two methods to identify and reject installs driven by click spam.

  • Hyper engagement: disqualifies high-frequency clicks, or engagements that we see too many duplicates of.
  • Distribution outlier: rejects attributions based on their click-to-install-time distribution.

Hyper engagement

The first step Adjust performs is eliminating high-frequency click spam. This occurs when fraudsters repeatedly send clicks on behalf of users in recurring intervals. In doing so, they hope to produce a “last click” that is relatively close to the install.

When an install occurs, Adjust checks all eligible clicks within the relevant attribution window. If we recognize any high-volume click patterns, we remove the clicks from consideration for attribution. This allows us to correctly attribute the install to the next legitimate click or as an organic user.

Distribution Outlier

To develop our distribution outlier real-time filtering method, we reviewed statistical data and analyzed legitimate and actual fraudulent activity. This research determined that over 85% of installs from legitimate traffic occur within the first hour after the ad click. This indicates a strong correlation between the ad click and the install time.

With click spam, however, there is no such time correlation between the click and install. This is because click spam poaches organic users or steals attribution from legitimate sources; therefore, the user never clicked - or even saw - the ad in the first place. Consequently, the click-to-install-time (CTIT) distribution on campaigns affected by click spam is spread out across the entire attribution window.

A visual representation of how the click-to-install-time distribution on campaigns affected by click spam is spread out across the entire attribution window.

As a result, Adjust will always prioritize engagements that show strong user intent. We determine this by analyzing the click-to-install-time distribution. If the majority of conversions occur 60 minutes or more after the click-time, Adjust will begin to reject attribution to clicks. These are then reported as distribution outliers, and those installs are attributed to the next eligible source or as organic traffic.

Flagging non-fraudulent traffic

Due to how Adjust analyzes user intent, there can be cases where non-fraudulent traffic is flagged for distribution outliers. For example, if you have a particularly high click-through-rate on your ads but low conversions, Adjust would analyze the click-to-install time distribution and potentially begin to reject attribution to clicks. This is because Adjust prioritizes engagements that show strong user intent, and our Distribution Outlier filter will act to prevent organic traffic from being misattributed.

If you notice rejections occuring, you may want to check that there are no technical issues delaying or obstructing users in their conversion journey from impression to click and install. In addition to resolving abnormalities in your data and providing consistency with the partners you work with, this also helps improve the overall user experience. Sometimes there is no technical issue, but instead it is the ad design which drives the high click-through-rate.

Blocking fraudulent activity

In addition to disregarding sources of fraudulent activity, Adjust uses the Blocklist API to block network-level links used by fraudulent actors. If a link is frequently associated with intentionally fraudulent activity, Adjust adds it to the blocklist. When a link is blocklisted, Adjust does not measure further interactions with it. Existing engagements and attribution remain associated to the blocklisted link.

Important:
Blocklisting is not a substitute for fraud prevention filtering. You must clarify your needs regarding indicators of click spam with your partners.

If you have any questions, reach out to your Adjust representatives.

Distribution modeling data in reports

Attributions rejected for click spamming will appear in your reports:

  • Rejected Installs: Too Many Engagements (RI TME)
  • Rejected Installs: Distribution Outlier (RI DO)

Reattributions rejected for click spamming will appear in your reports:

  • Rejected Reattributions: Too Many Engagements (RR TME)
  • Rejected Reattributions: Distribution Outlier (RR DO)
Note:
For installs where attribution was rejected for click spamming, Adjust will either reward attribution to the authentic source found through Adjust’s attribution methodology, or, if no other source is found, to Organic.