Ben Trenda, CEO with Are You A Human shares his programmatic views
PA.O: Tell us more about the differences between “robot” behavior and “human” behavior. Where do they differ? Where do they overlap?
Ben Trenda: Almost all sophisticated bots mimic human behavior, but bots are generally written to do one thing very well. For example, posting links in forums to improve search rankings. So on some sites bots will appear to behave human, but what they don’t do is act like a person in every situation, on every site. It’s not simply “is this behavior human?”, but rather “how does this user’s behavior compare to human behavior both on this site and others, and also how have we seen this user behave in other scenarios?”
PA.O: What are the challenges faced in building “a better mousetrap” for catching ad fraud?
Ben Trenda: We aren’t building a mousetrap at all. We don’t look for bots, but instead search for real humans. It’s similar to border security. You don’t make a list of every single person in the world not allowed in your country, instead you make a list of citizens allowed in, and give them a passport. We’re doing the same for real humans online. Also, thinking about stopping ad fraud bots misses the scale of the problem. Most wasted ad spend does not come from malicious ad fraud, people purposely driving fake traffic to a fake site. The big problem is that the majority of all traffic on the internet is generated by bots. There are millions of bots crawling the web for a million reasons, good and bad. Data scrapers, Google’s crawlers, spam, content piracy, data aggregators, classification, etc. None of these bots care about ad dollars, but they hit millions of pages and register millions of impressions. Wasted ad spend is just a side effect. So even if we stop ad fraud of all kinds, there will still be billions of dollars wasted showing ads to all the bots moving around the internet for their own reasons. What we’re doing is giving websites (or advertisers) the ability to know if a visitor to there site is a real person or not. There are many ways that sites use that data, deciding if an advertiser should pay for an impression is just one of them.
PA.O: When distinguishing between human vs. robot, is there a difference to note in monitoring B2B vs. B2C consumers? Across devices – i.e. mobile vs. tablet vs. desktop?
Ben Trenda: Humans pretty much behave like humans. We do of course see differences between mobile and desktop since the input device is different.
PA.O: How can one assure that they are never eliminating human traffic? Is there a threshold of humans lost for every human verified?
Ben Trenda: There aren’t millions of new humans coming online every day. We haven’t whitelisted every user in the country yet, but once we do the list will be fairly stable. Users will move in and out of the list as they accidentally install malware and stop behaving human, but once they clean their device we can see them behaving human again and add them back to the list. Compare that to blacklisting, where you see literally millions of new bots created every day. Bots only last for a few days so by the time you blacklist them your list is already out of date. So, while it is possible you may miss a few humans by taking a whilelisting approach, taking the blacklist approach guarantees that you will always miss millions of bots. Its an endless game of whack-a-mole.