BORIS MOUZYKANTSKII: Germany is an extremely important market for us. At least from my limited interaction here, you tend to get into technical discussions very quickly and having some German engineers on our team will help us to serve this market better. Historically, as a company, we worked closely with Adternity for many years. They were focusing on their UI solutions and we were building the back end and we still have many clients whom we serve directly so there was a lot of overlap this way.
What we realized going forward is that custom solutions will be more and more in demand not by traditional ad tech, venture-funded companies, but more established publishers or advertisers who are really not that interested in building UI themselves and generally want something ready. It can be very custom, but they wouldn’t want to do [the heavy lifting] themselves.
Can you define “custom” solution? Is this the same as a customized algorithm?
The promise of real-time bidding is mind-blowing. It basically says that for every single impression, (a machine can) decide what do you think about the person who is about to see this impression and what do you want to say to this person and how much do they want to pay for the opportunity to talk to this person. Now we’re at the very first sort of stage of this evolution, and it’s similar to when they started commercializing electricity in the early 20th century. The first [product was] lightbulbs and they very quickly realized, “I can make a fridge run and a vacuum clean.” And it [spread] everywhere. In the same way, RTB can completely change marketing, but for now it’s a media-buying system.
Now that you have Adternity, is IPONWEB building an ad stack?
The word “stack” is sort of overused a bit. We have a set of components and the best way to describe it is those components are void of business logic. They are pure engineering pieces and we can combine them together to create something unusual or usual. It’s just a question of what the customer needs. I wouldn’t call it a “stack,” because in that respect we’re different. There are many cases where all you need is a solution for a very well-defined, existing problem and a complete stack of vendors like Google just fit because they understand the problem, they found a solution, and it fits. Now, through the nature of RTB interaction, most of the problems advertisers face are extremely custom [and require customized algorithms and modeling.] The way an airline communicates with a potential traveler is very different than the way a bank communicates with a person. The problem is not solved yet.
You were saying RTB will filter into other aspects of the marketing “ecosystem?” Isn’t this already happening?
To a large extent, RTB is still a light source. It’s changing and actually quite fast and now you have the data connections. The advertisers are not yet in a position where they do really [customized ad serves]… There is alot of decisioning and very many impression opportunities, [but] the nature of algorithms we run are still relatively primitive, they’re not custom to the advertiser to a large extent, and we see that as a large opportunity.
How would you describe the evolution of IPONWEB? Your heritage was really building ad servers.
We started in this space many years ago almost by accident. RightMedia found us and asked us to help them with testing. Back then, they had about three people in their technical team. It was a really small startup and we were helping them test their ad server. As time went by, we got more and more involved. In 2007, when they sold to Yahoo, we had 50-odd engineers working with them on the RightMedia account. But when Yahoo acquired RightMedia, we realized our biggest client was about to disappear and we got a nice transitioning contract and they almost bootstrapped our expansion and activity.
We thought, “What can we do?” and we realized, “Well, we know the space really well.” We didn’t have any IP back then because work for RightMedia was work for hire and so we thought, “Let’s build some set of components or tools and then go and build custom ad serving systems for other clients.” Probably, if I had more business sense, I would have shot this idea down because back in 2008, it made very little sense. By late 2009, AdX opened up their RTB product and suddenly, pretty much every technology stack in the space needs to be adjusted, rewritten and there were just totally new ways of trading ads, such as SSPs, DSPs – all of this appeared on the back of this technology innovation. We got a lot of interesting business back then…and as far as I’m concerned, if Google was delayed two or three years, I probably wouldn’t have a business.
You’ve just teamed up with Ghostery to detect fraud in the RTB ecosystem. What spurred this besides advertiser demand?
One idea we seriously thought about is if we’re bringing in all those smaller supply sources, then we should probably take a little bit of care to determine is it human traffic, bot traffic? We do impression-anomaly detection, which is a polite way of saying if we suspect traffic is non-human, we just don’t pass it on. And then the next iteration on this idea was, “OK – let’s deal with human traffic,” but what if a publisher, SSP or somewhere in between presents an opportunity to an advertiser on Expedia.com and it has nothing to do with Expedia?
This could be a trafficking error or malicious intent, but how do you catch it? The conversation with Ghostery was they effectively have a panel of users sitting in many, many browsers and they know what those users are seeing. The idea was quite simply to say, “We know this opportunity was offered to us” and most likely, it was offered to many other parties. And very likely it was fulfilled and there was a very good chance one of the Ghostery users saw that ad as a result of fulfilling this opportunity.
If we can look and try and match together those two events, then we can identify situations when the opportunity was presented from one domain and presented on another one. If it happens more than once, we can be reasonably sure that something is not quite working. And it’s a relatively robust way to measure this because you know what was in the bid offer and you know what sits in the browser, because Ghostery’s technology verifies that.
What’s next? What are your plans for Adternity?
Because we had joint clients for many years, on the tech front things are pretty much the same. We probably will invest a little more into their technology stack, but from a client perspective, they used to have two contracts and will now have one, but roughly things are similar with teams and divisional responsibility.