This recommendation is based on the simple concept that even if an ad is 100% viewable, if it was caused to load by a bot, it is still fraud, and useless for advertisers, who thought their ad was
shown to a prospective human customer.
Bots have been around for as long as the Internet, first known as viruses and malware that infected people’s home computers. They stole personal
information and visited websites, per the direction of their botmasters, to create fake traffic, ad impressions, and clicks.
But the days of relying on unsuspecting humans to accidentally
click and download malware onto their PCs are limited. Modern-day botmasters can “spin up” millions of headless browsers in data centers to commit ad fraud on a far larger scale than could
be done with individual PCs. These headless browsers were designed to help developers pressure test websites before launching them. The bad guys use millions of these fake browsers to simulate humans
visiting websites, therefore generating ad impressions, video views, and clicks.
advertisement
advertisement
The more sophisticated bots have been observed to create fake mouse movements and page scrolling. They can add
items to shopping carts, deliberately abandon them, collect cookies, and wait for retargeting to follow them to websites owned by the bad guys. When the ad impression is served on those sites, the bad
guys make off with the money.
These bots are irresistible because they generate the lift in traffic that every publisher lusts for. They create ad impressions at very low cost, something that
media buyers lust for -- to give their clients more impressions at lower average cost. And they generate the click-throughs that are thought to mean engagement and actions leading to conversions that
advertisers love.
If you analyze more deeply and carefully, these ad impressions, visits and clicks all end with no conversions. I don’t mean just ecommerce sales; I mean any conversion
event that brand advertisers could measure on the path towards sales. This is because all these actions are caused by bots.
So, the bot problem and NHT (non-human traffic) need to be solved
first. Some commercial anti-fraud vendors are doing what they are required to do: screen against blacklists provided by industry associations. Unfortunately those lists contain a fraction of the
approximately 10,000 known bots that have been observed in the wild.
But the bad-guy bots are NOT these. Bad-guy bots do not declare themselves honestly in the user agent, as Googlebot,
Bingbot, Facebookbot, etc. do. Bad-guy bots disguise themselves as Internet Explorer, mobile Safari, and every other flavor of browser that humans would use.
Some published numbers suggest
that the bot problem is only 1% - 3% of total traffic. That would be true if you only counted search engine crawlers and known bots. It would be a dramatic underestimate if you are looking for fraud
bots.
What advertisers and agencies should do is immediately leverage the anti-fraud technology vendors to help them identify highly suspicious bot activity using big data analytics, machine
learning, and heuristics around network traffic. Using this kind of “broad strokes” detection, the most obvious bots can be caught. The bottom-most decile (10%) should be lopped off, by
blacklisting those sites that loaded the impressions or sent the traffic. This will have a positive business impact even while viewability is still being worked out.