Analyzing Lag Time Between Click & Install

As a marketer, you rely on publishers to not only drive traffic to your app, but provide quality and legitimate traffic. Fraudulent or low-quality traffic can have negative impacts on campaigns, which can be difficult to identify and filter.

TUNE’s Fraud Analysis reports make it easier for marketers to analyze their paid traffic and highlights characteristics for potential fraud. One of these reports is the lag time variation report, which displays the time difference from click to install.

This article is part of our Popular Features series.

Lag Time Variation Report

Trending the pattern of the installs over time shows advertising in action; a rise in installs shortly after the click, with a decay in the install rate over time as the ad loses recency and the user moves on to other things. The pattern of installs can also be used as an indicator of two categories of click-fraud that may represent as much 70% of the overall fraud.

The specific metrics used to help analyze the data  are in place to help profile the install patterns appearing. These thresholds themselves do not directly identify fraud sources. For example, these metrics should not be interpreted as a 9 second install is fraudulent, and an 11 second install is valid.

Click spamming, ad stacking, and other related techniques used to cannibalize organic traffic can be identified by a relatively flat install curve over time. Click injection can be identified by a high number of installs happening far faster than is typical (for your app).

Click-Injection Fraud

The percentage of traffic that comes in under 10 seconds for a given publisher, campaign, sub-publisher, or country can be compared with the percentage of traffic that comes in under 10 seconds overall for your app. For example, if the overall traffic has 2.5% and one of the Partners has 17% of installs arriving under 10 seconds that would be an indication of fraud possible within that traffic.

Using the Lag Time Variation report to drill down to the sub-publisher level may show one sub-publisher with 22% installs arriving under 10 seconds and another with 3% arriving under 10 seconds. In this example, the sub-publisher with 22% would show indications of fraud traffic across that entire traffic source, and the recommendation would be to review (or pause and review) that entire traffic source.

Click-Spamming Fraud

Linear patterns across most of the installs would be an indication of click spamming in action. When most of the traffic follows a relatively flat install curve, the advertising did not drive most of the installs. In the graphic below, there is a small peak near the ad click indicating that the ad did have impact, however the overall install curve is flat indicating that most of the installs were not driven by the ad.

Drilling Down Into the Lag Time Variation Report

By default, the lag time variation report shows the number of total installs, the percentage of installs under 10 seconds, and the mean time to install grouped by source (i.e. your advertising partners). You can further drill down into the report by making use of various groupings, metrics and filters.

Groups

The Groups tab enables you to select different variables to group installs by. We recommend always including the “Partner” field so you can easily keep track of which partner drove the installs regardless of how you slice the data.

Metrics

The Metrics tab enables you to select one or multiple metrics to be included in the report; installs, % installs under 10 sec., and mean time to install are included in the report by default.

Filters

The Filters tab enables you to further refine the report by displaying on the data that matches the filters you apply.

Things to Keep In Mind

  • In some cases, TUNE must recreate the actual click to use for attribution. As TUNE recreates these clicks simultaneously with the install itself (0 lag time), this report does not include TUNE-generated clicks.
  • The Installs by Second After Click graphic only shows installs that occur within the first hour of the click; i.e. axis is limited to 3600 seconds.
  • The date range at the top represents the install date range, not the click date range.
  • Data in this report only includes installs where the click happened within 7 days prior to the install.

No Comments

Leave a reply