How do you detect traffic blending?

Quality that drifts down gradually as volume from a source grows. Sharp differences in conversion between sub-IDs inside one source. Volume that rises exactly when asked, with no change in price. Revenue per click falling while click volume holds steady. Deductions at month end that concentrate on one source.

How do you prevent traffic blending?

Demand sub-source transparency and pass a Sub ID for each one. Evaluate at the most granular level available, using Keyword-level reporting and hourly data. Set a minimum quality score per sub-source, not per source. Scale sources slowly; sudden extra supply is often filler.

Traffic blending: how it works, warning signs and defences

Traffic blending

Low-quality or fake visits are mixed into a stream of good ones so the average still looks acceptable.

The short answer, from The Arbitrage Desk fraud map

How it works

Imagine a milk seller adding water: a little goes unnoticed, and each litre of water is sold at the price of milk. Traffic blending does the same with visitors. A source with genuinely good traffic adds cheap bot, incentivised or off-target traffic until the blended quality sits just above the level that triggers alarms.

Blending works because feeds and buyers often judge a source by averages. The remedy is to stop looking at averages: split the source by sub-source, hour, device and country and the watered portions show up as pockets with no conversions.

Who pays for it

Advertisers pay full price for the diluted portion. The arbitrageur buying the blend pays for visitors that earn nothing. Over time smart pricing lowers the value of every click from the account, including the good ones.

Who does it, and why

Traffic networks and sub-publishers who are paid per visit or per click and can buy filler for less than they are paid. The margin is the price difference between good and bad traffic.

Warning signs

Quality that drifts down gradually as volume from a source grows.
Sharp differences in conversion between sub-IDs inside one source.
Volume that rises exactly when asked, with no change in price.
Revenue per click falling while click volume holds steady.
Deductions at month end that concentrate on one source.

Defences

Demand sub-source transparency and pass a Sub ID for each one.
Evaluate at the most granular level available, using keyword-level reporting and hourly data.
Set a minimum quality score per sub-source, not per source.
Scale sources slowly; sudden extra supply is often filler.

An example

Illustrative: a source delivers 10,000 clicks. 7,000 are real and earn $0.40 each; 3,000 are filler and earn nothing after review. Revenue is 7,000 × $0.40 = $2,800, or $0.28 per click on average. The buyer paid $0.25 a click, $2,500 in total, and believes the source is profitable at a 12% margin. Removing the filler would have cost 7,000 × $0.25 = $1,750 for the same $2,800: a 60% margin. The blend hid a $750 loss inside a "profitable" source.

Documented cases

Spyware clicks inside Yahoo's syndication chain (2006): Syndication fraud: adware generating paid clicks through layers of search partners