A Peek Inside Google’s Click Fraud Detection Systems

Matt Cutts points to a  very interesting report on Google click detection and evalutaion methodology. The report is by Dr Alexander Tuzhilin, who “was asked to evaluate Google’s invalid click detection efforts and to conclude whether these efforts are reasonable or not”.

In order to do so, he spent time at Google, evaluated internal documents, interviewed employees and was given demonstrations of their systems.

If you don’t want to read the whole thing, he concludes: their efforts are reasonable.

Some highlights:

  • It makes no business sense to go after extra revenues (i.e. benefit from click fraud) if Google loses the trust of advertisers.
  • Google has built the following four lines of defense against click fraud – pre-filtering, online filtering, offline filtering (automated) & offline filtering (manual)
  • They use anomaly based, rule based and classifier based systems
  • Google defines click-spam as “Click Spam is any kind of click received from a Cost-Per-Click advertising engine that is generated artificially through human or technological means with the sole purpose of creating a debiting click, resulting in zero possibility for a conversion to occur”

There’s a lot of interesting stuff in here that hasn’t, to my knowledge, been detailed online up until now. Reading….

Leave a Reply