---
"title": "Best Practices for Combating Click Fraud [Data Series, Part 2]"
"date": "2017-03-15T18:11:19-04:00"
"url": "https://www.criteo.com/blog/best-practices-combating-click-fraud-data-series-part-2/"
"image": "https://www.criteo.com/wp-content/uploads/2017/07/Blog-2017March05-Mouse-Click-Fraud-720x400-min.png"
"author": "Brooke Freeman"
"category": "Commerce Media"
---

In our second post of this series, we’re discussing ways to manage click fraud.




## **Approach #1 – User agent**




The simplest approach to monitoring bot clicks is to block clicks generated by self-proclaiming bots. Some bots scour the web to collect information. These bots often purposely communicate to other parties that they are bots through their user agent, a means by which a web browser tells a website information about itself. The common convention is to include the term “bot” in your user agent when employing a bot. Other terms that may be useful to monitor include “crawler” and “spider”.




## **Approach #2 – Click-through rate**




One of the most effective approaches to monitoring bot clicks is using click-through rate. Humans typically click on less than 5% of display ad impressions they observe. If you notice an IP address clicking on ads an inhumanly large percentage of the time, that’s a clear indicator that the IP has been compromised by a bot. However, humans may open up a page with a few ads, click an ad, then end their browsing session. Because of this, in order to confidently use click-through rate for bot detection, some minimum threshold of clicks per time period must first be observed.




## **Approach #3 – Frequency**




Another commonly used bot monitoring approach is actions — such as clicks — per time period. This is useful for covering cases where bots mimic human click-through rates but instead rely on volume to create sizable fraud. Humans typically click on less than 10 ads in a given minute. If you notice a cookie clicking many times per minute, that’s a clear indicator that the cookie has been compromised by a bot. Keep in mind that IPs may comprise a large number of devices, so attempts to count frequency by IP would need to be done conservatively.




*In part three of the series, we’ll be looking at how only about half of display ads are ever viewed.*




4-Part Data Series:  

[The Importance of Data Fidelity in Advertising [Part 1] ](http://www.criteo.com/blog/importance-data-fidelity-advertising-data-series-part-1/)  

[Best Practices for Combating Click Fraud [Part 2]](http://www.criteo.com/blog/best-practices-combating-click-fraud-data-series-part-2/)  

[Why to Incorporate Ad Viewability Into Your KPIs [Part 3]](http://www.criteo.com/blog/incorporate-ad-viewability-kpis-data-series-part-3/)  

[The Relationship Between Attribution and Performance Marketing [Part 4]](http://www.criteo.com/blog/relationship-attribution-performance-marketing-data-series-part-4/)