AdExchanger: Why get into ad fraud detection?
JACOB LOVELESS: We had customers coming to us late last year saying that one of their big problems is advertising fraud. Advertising is so critical to their business that literally every dollar of efficiency is the difference between success and failure.
At the time, I actually supposed that ad fraud was not as serious as people thought. And then we got into it and were like, “Okay, actually, this is a real problem. This is a serious problem.”
We looked around at what everybody else does in the market, and the problem we saw was that identifying bots is fairly easy. I mean, I’ll put “easy” in quotes, but identifying whether something is a bot or is coming from a data center – lots of solutions can do that. But the only way to really solve the problem of ad fraud is to stop showing ads to fraudulent people. You can’t solve it retroactively. Advertising is pay per click, not per view.
What’s your solution?
When the bad guys show up, we send what’s called a redirect that effectively moves them to a sub-path of the website.
With that subtle redirect, all of the trackers a customer has set up fire. The Facebook tracker fires, Google Ads fires, Criteo, TikTok – all the things that they use to do retargeting. Then we can set their audience definitions not to show ads to anybody who has visited this site’s sub-path.
We take the most powerful part of an ad network, which is the retargeting engine, and let it work for you. Effectively your retargeting is identifying bots so that they never see ads.
What’s the false positive rate for the users you channel to a sub-path?
Bot protection tech is pretty robust. We’re sitting on top of a couple of the top bot protection engines that you know. The false positive rate on that stuff is low, sub 1%.
Where it gets more sophisticated is when you start bringing in IP-quality information, and for this you need to go out and license databases that update in real time.
What’s neat about our signal system is that we’re effectively the front door of a website. We see all of a customer’s first-party cookies and headers and we know what users did during a session, like whether they added something to their cart or exhibited buying behavior. We can take the IP-quality databases that, frankly, have a high false positive rate – probably in the 4% to 5% range – and combine them across the Edgemesh customer base.
That way we can know if an IP address has a history of fraud, for example, but actually has done a checkout or an add-to-cart somewhere. In that case, perhaps we wouldn’t include it in the redirect script. This type of intelligence knocks down the false positive rate and also gives us some signals that don’t exist in the licensed databases.
We bring that information into the system so every customer gets protection.
Do you plan to go further into advertising analytics?
We will eventually get there. The next thing for us on the analytics and ad attribution side will be to build a replacement for Google Analytics. They’re deprecating the original Universal Analytics product and, unfortunately, the new GA4 for ecommerce companies is a really tough sell. A lot of things don’t work.
Commerce has additional dimensions that need to be in the system. You need to know things like add to cart, the checkout rate or whether a purchase price includes shipping or not. In Universal Analytics, this stuff was all done with a product called Enhanced Ecommerce that was widely implemented.
But now, things that we used to take for granted – like ecommerce conversion rates – are not in the system. For example, there was a section where ecommerce customers could see ecommerce rate, online conversion rate, average order value and average session value, all with a click of a button. GA4 does not have any of that stuff today.
The other big thing for ecommerce in particular is that historical data doesn’t come with you. Year-over-year comparisons and seasonality forecasts are critical for a commerce business, and because it was free for all Google Analytics customers, people didn’t invest in another system to house that info, even though its crucial for the business.
When Universal Analytics goes away, your data starts over with no ability to compare or look back.
If the data was so critical, why was everyone seemingly so fully invested in Google Analytics?
There was no market for a competitor. What investor is going to back tech when the market leader is Google offering something for free? The answer is nobody. That’s a tough sell.
At the high end, you have Adobe and other services working with enterprise brands, but the mid-market and down is universally on Universal Analytics.
The main reason GA4 doesn’t log IP data is because of privacy concerns, but Edgemesh sees and uses IP addresses. If you were to delve further into advertising you’d need to start thinking about consent and things like legitimate interest, right?
We are allowed to use IP addresses, just like a security company can use that data to make implementations. And we don’t really collect the rest of the stuff, because we don’t have a use for it.
If we were to go into the analytics space, then yes, we would absolutely need to collect that data and become a data processor, which is fine. That’s not a thing we’re worried about.
We believe that products need to deliver value. Dashboards are where data goes to die. So, if we were to build an analytics product, we would want the analytics to do something for the customer automatically.
How do you mean?
For example, some products sell out very quickly. Should a seller increase prices as inventory levels drop? I can tell you that they should certainly decrease the advertising spend on that campaign. If you just launched a new shoe and it’s going to sell out in the next eight hours, you probably don’t need to keep pumping money into advertising campaigns.
We’ll keep analytics close to the problems our customers have and the feedback they give us. Our new first advertising solution was directly related to customer feedback on fraud.
We’re not just trying to get the data, but to turn that data into value.
This interview has been edited and condensed.