A New AppNexus? CEO O'Kelley Talks About Ch-Ch-Changes On Eve Of Expected IPO

okelley-oct15AppNexus has long been the grand poobah of indie programmatic media platforms. But lately it has begun to look like a company under siege.

There's the fraud issue, which became an albatross after competitors Rubicon Project and OpenX cleaned up two years ago while AppNexus continued to allow blind impression resale. It has since fixed that problem.

Meanwhile the rise of walled gardens in mobile has cut off swaths of essential supply such as Facebook's mobile news feed. A third challenge is key supply partner Microsoft's decision to outsource much of its ad sales to AOL. That smarted, but AppNexus preserves a still-large chunk of the business.

Even so, AppNexus looks strong. It has continued to invest in its tools both on the buy and sell sides as it inches toward a likely IPO in the first half of 2016. Its valuation on the secondary market hovered above $1 billion earlier this year, according to sources, making it one of the few ad tech industry unicorns.

AdExchanger checked in with CEO Brian O'Kelley.

AdExchanger: In June you said upward of 40% of your supply had been deemed fraudulent and removed from the AppNexus marketplace. More recently in July, you put the number as high as 65%. How did you get to such a large number?

BRIAN O’KELLEY: We should distinguish what percentage is fraud, versus what percentage is invalid. There's a big difference. What we've seen is a given impression can bounce around between multiple players. That's one bad impression we saw 10 times. How do you count that?

The first thing we did was change the policy not to allow impression resale. Just doing that compressed the number of actual impressions we saw dramatically. That was nice because we then knew the source of the traffic. Everyone's accountable for the traffic they bring to the platform. It compresses the supply chain.

Secondly, we enforced supply-chain transparency. If you are an SSP, we need to know the actual name of the publisher you work with.

That eliminated a ton of impression volume, but I don't think it actually eliminated that many underlying impressions. So volume on our platform compressed dramatically, but I'm guessing the amount of real traffic impacted was relatively small. Only 3% of all spend went to impressions that left the platform.

What percentage of your fees were removed?

Most of our fees come from transacting on net spend. We do get paid for ad-serving volume when things don't clear, but the charge is almost the same as the cost. So removing all of the non-transacted impressions also removes all the non-transacted cost. So we weren't making money from that. The percentage of fees we lost was not zero, but it was immaterial.

How do you answer buyers who think you waited too long to clean things up?

There are some mixed emotions about us doing this. For a lot of the ad networks, the problem is the arrow points through the whole supply chain. We didn't make them buy those impressions. We didn't make them explicitly turn on the sellers. We just said, "We're open and this is between y'all." And now we're saying, "It's no longer between y'all, we're going to change the policy and make some rules here." It pissed a lot of people off because they were very happy with how things were.

If you're a buyer unhappy with having bought the traffic from AppNexus, it was you who put in the bid request. You decided you wanted to buy traffic from that domain, from a seller that we disclosed to you, and we made no argument for or against that impression. We've always been neutral. Saying, "You guys should've had a stronger policy," well OK. But we do have policies. We have always advised buyers on how to enforce quality if they want to. We have pre-bid integration with every major quality provider.

I ignore that in general. Haters gonna hate.

Does the timing have anything to do with your upcoming IPO? Does this help you make the case to public investors that you have a clean ecosystem?

Regarding the timing of this move, two things came together. One is we figured out how to fix it, and two is we wanted to build quality protection into the buy side. We're a buy-side player, and for those clients we had to build this tech anyway.

I think you're right that, for any public company that doesn't do a good job protecting against quality issues, there's risk. But I don't think it's directly connected. It has to do more with being a good actor and a good company.

Speaking of your buy-side business, what do agencies and marketers contribute to your revenue percentagewise?

We're not disclosing revenue information.

Do you call directly on marketers?

We don't typically talk to marketers without the support of agencies. There are some marketers we do. One example is Wayfair. They don't work with agencies, so they use us.

It's much more common for large marketers to work with us via the agency or via the trading desk.

Does WPP Group's 15% ownership stake make it hard to get marketer business through rival agency groups?

No. We have two large financial services clients who are with Publicis shops, and we have strong technology relationships with them. The thing about holding companies is they are distinct from operating agencies. [Publicis CEO] Maurice Levy might say, "I don't want to give business to AppNexus because WPP owns 15%." But if we're the best fit for a marketer, the agency is going to work with AppNexus.

Basically, if Martin wants us exclusive, he can write a check for the 85% of us he doesn't already own. I wouldn't have taken the money if it had any impact on my ability to do business with other holding companies.

We have a global partnership with Omnicom, we have a big partnership with Aegis/Dentsu. We cannot compromise those.

Let's talk about your publisher business. What's the whole vision there, and how does it challenge Google?

We call it full-stack, which is ad serving with built-in pre-bid RTB and world-class forecasting and analytics from Yieldex. Many of the largest DFP customers use Yieldex for forecasting because it's much more accurate. If you use accurate forecasting you can make a real-time trade-off between a programmatic impression and a guaranteed impression.

To take a concrete example, if you've got a guaranteed buy where you want to spend $100,000 over a month for males on the home page, I can still fulfill that, and let's say a big marketer like P&G comes in to cherry-pick an impression for a $20 CPM. Yieldex gave us the forecasting to make that decision more accurately, and so we see significant yield lift over the equivalent setup in DFP.

If you look at what DFP has done with enhanced dynamic allocation, you can only do that with DoubleClick AdEx, and you can only use their built-in forecasting. With us you get all the demand sources, like you do with pre-bid, and you get Yieldex forecasting. It's a much better technology.

In terms of progress, we are ahead of schedule. We won a major deal against DFP in the Nordics with Schibsted.

What about walled gardens? There's more supply sitting behind proprietary ad stacks than ever, including those operated by Google and Facebook. How do you combat that trend of isolated pockets of media, especially in mobile?

Right now, FBX is fully open, and we don't have access to the Facebook mobile news feed, which is obviously a huge chunk of inventory. Google is fully open today, but is planning to close YouTube in January. Yahoo is open as far as I know. Yahoo doesn't have a lot of owned and operated. The only way for BrightRoll to be a successful platform is to be open. I would expect Yahoo to be aggressively open. Twitter is interesting, because with MoPub they've made "open" a cornerstone of their strategy.

And don't forget, there are so many properties out there. There's Pinterest, Snapchat, LinkedIn.

You sound pretty optimistic.

I'm not sure that the two chunks of inventory that are closed are enough to say the whole industry is going closed. And I'm not convinced those two are going to stay closed. If you're P&G, you can say to Facebook, "Look I'm writing you a check for $100 million a year, and you're telling me I can't bring my full [tool set] to bear. I'm not down with that."

If you talk to the Facebook team, they're not religious about being closed. In fact, both Google and Facebook are open by nature. The biggest issue right now is going to be the privacy and identity concerns around how they leverage their own data. It's not open inventory vs. closed, it's open user data vs. closed.

How bad is Microsoft's AOL deal for AppNexus? Microsoft is your key supply relationship.

We keep the globe from a technology perspective. In nine markets, if you do any programmatic deal with Microsoft, AOL is the programmatic sales team and we are the programmatic technology.

In 10 more markets, we're both the programmatic sales force and the programmatic technology. And in those markets sales are going all programmatic. The initial results are very good, because the net profit increases.

What are your feelings about AOL, since they took a chunk of the Microsoft business? They also grabbed Millennial Media, which was once a key partner of yours at one time.

We have to have a strong partnership, because we share the same key partner.

Do you have the video tech you need right now?

Yes.

Do you have the video supply you need?

Most of our supply comes from the SSPs, so LiveRail, Adap.tv, Google and others. We don't have an SSP product in market for video yet. We will next year.

On the topic of ad blocking, is site serving a legit way to address the trend? Can Open Ad Stream help?

We've been doing that forever. OAS was one of the first ad servers to do first-party ad serving. And there's a huge chunk of what they call hosted ad serving that's going on today.

But I think ad blockers are pretty smart. Just the fact that it's being served from the same domain will not solve the problem.

Why not?

By trying to do site serving you've got to move to first-party creative delivery, and by doing that you actually fix the user-experience problem. Just beginning that conversation starts to change how ad delivery works, which starts to change the user experience, which starts to fix the problem. I think it's going to be a great driver for important changes in how ads are served.

What about going after the ad blockers directly?

Do we want to be in an arms race? It seems dangerous. I'd rather find a path that respects the concerns about user experience and privacy. Do we prefer the government fixing it, or fixing it ourselves? This is even worse than the government. This is the people. People are really hard to lobby.

It's not just an ad tech problem, it's not just a publisher problem, it's a consumer problem. Outside the business press, there's not much content that's funded by subscription.

Will you launch a CMS at some point?

I don't know, but I think we'll be plugging into CMSs, and I think CMSs will be plugging into Internet tech platforms. It's going to be very connected, because the content needs to be contextually relevant.

The ad-delivery conversation is getting simpler. Because consumers won't accept crappy experiences, the options will be limited. All the sliding ads, takeovers, none of that is going to work because consumers won't tolerate it.

3 Comments

  1. John Musgrave

    It's hard to fathom how 65% of traffic could have only generated 3% of spend flowing through Appnexus's system. Were it true, it obviously implies that the other 35% of inventory (the non-fradulent inventory) was generating the other 97% of spend. If that's the case, that means that the CPM of the non-fraudulent inventory was 60x the CPM of the fraudulent inventory which would seem to indicate that buyers were already able to identify and avoid fraudulent inventory. That seems incredibly unlikely to have been the case given the uproar from buyers about the quality of the traffic within Appnexus which either means that Appnexus didn't actually cut 65% of traffic or that it had to have more than a 3% impact on spend.

    Reply
  2. It's hard to fathom how 65% of traffic could have only generated 3% of spend flowing through Appnexus's system. Were it true, it obviously implies that the other 35% of inventory (the non-fradulent inventory) was generating the other 97% of spend. If that's the case, that means that the CPM of the non-fraudulent inventory was 60x the CPM of the fraudulent inventory which would seem to indicate that buyers were already able to identify and avoid fraudulent inventory. That seems incredibly unlikely to have been the case given the uproar from buyers about the quality of the traffic within Appnexus which either means that Appnexus didn't actually cut 65% of traffic or that it had to have more than a 3% impact on spend.

    Reply
  3. "The first thing we did was change the policy not to allow impression resale. Just doing that compressed the number of actual impressions we saw dramatically."

    This is a key point Brian is making. It is exactly the kind of thing that IAB should be speaking about as well. Redirects, ease of getting tag payloads in the eco-system at scale and the fact that ad networks pocket most of the ad fraud money are IMO the most important factors to deal with at the industry level.

    It is noteworthy that AppNexus have been and continues to be so bold in regards to this matter, both in terms of talk and walk. It is easy to criticize virtually any company that is involved in large scale internet advertising, but it's not always easy to say that one is behaving in a courageous manner. We saw that the companies that had listed so far, took a different approach.

    Reply

Add a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>