At the most recent Game Developers Conference (GDC) in San Francisco, I had the pleasure of moderating a panel discussion about the dawn of in-app header bidding—the digital advertising technique turning mobile game ad monetization on its head. In particular, I was joined by five of mobile gaming’s most experienced monetization experts: Glu’s Director of Ad Monetization Brian Kealer, AdColony’s EVP Global Publishing, Partner and Account Management David Pokress, SEGA’s Senior Director of Ad Monetization Adam Carey, Pixelberry Studios’s Head of Marketing and Ad Monetization Filippo De Rose, and MobilityWare’s Lead Analyst Monetization Michael Ruffin.
We discussed how in-app header bidding changes the way ad monetization teams work, why mobile game publishers should evaluate each in-app header bidding solution they come across, how mobile ad networks view the in-app header bidding revolution, why network SDKs play an important role in the success of in-app header bidding, and much more.
What follows is a lightly edited transcript of the conversation.
Christy Wong-Taylor, Fyber: If header bidding solutions enable you guys to avoid the waterfall, then obviously there’s going to be an operational impact. So can you tell us a little bit about your current day-to-day? What is it like for people on your team that actually manage ads and, in particular, how often they make changes to the waterfall?
Adam Carey, SEGA: Well we really don’t change the waterfall nearly enough. A lot of our traffic is driven by a few titles, so we tend to focus on that. It’s way too difficult just to keep moving the waterfall around.
But I worked on the desktop side for 19 years, so we deployed header bidding. Big reveal: there is no header in the app, so we should stop calling it header bidding, please?
On an everyday basis, we’re not moving the waterfall nearly enough. I’m looking forward to the impact from header bidding I saw on desktop. We saw about a 20% lift in yield, which was fantastic. Some of that was the operational efficiency—getting the right bid in. But a lot of it was it opened up brand new ad partners.
Michael Ruffin, MobilityWare: We very much operate on a waterfall system. We built our own mediation stack, mainly because there were no mediation stacks when we came out with Solitaire. So we built our own, and we continue to iterate on it. My team is typically making changes once a week to the waterfall, depending on what we’re seeing. I think with header bidding, we’re not quite as bullish as a lot of other publishers out there. I think there are advantages for smaller publishers with unlocking more inventory. It’s going to be pockets of inventory that we’ll be able to monetize through header bidding.
Brian Kealer, Glu: At Glu, we have two teams split up on the ad monetization side. One is to grow supply, and work with the studio teams on adding additional placements to increase our ad impressions. Then we have another team that is dedicated just to pricing and yield. Those people are in their ad server or mediation layer all day, changing pricing, adding new line items into the ad server to hopefully increase the average eCPM, and trying to find the sweet spot for partners like AdColony to compete in the world of waterfalls today. That’s what we’re trying to get away from.
Filippo De Rose, Pixelberry Studios: We all want to optimize to get the best eCPM per country, but at the end of the day, the brutal honest truth is we optimize by the ad network that gives us the best dirty deal to be in first position.
Christy: After in-app header bidding scales, do you anticipate it changing your ad ops teams’ day-to-day?
Brian: I definitely do. I think header bidding will flatten the waterfall a little bit. It’ll increase efficiency while decreasing overhead on how much time we have to spend in our ad server, in terms of pricing.
Adam: Here’s the thing for me, I know there are good ad networks like AdColony. And sometimes they’re stuck down in the waterfall, and they’re not getting to the right person at the right time with the right ad, because maybe they weren’t bidding enough historically. In the Fyber FairBid world, AdColony will get that opportunity to actually deliver that ad to the right person.
We will see a good yield spike from this, whether it’s just through the efficiency of the actual mechanism. I can’t believe that we’re still using waterfalls. Again, coming from the desktop world, I have not heard the term “waterfall” for about eight years, and then I heard it like ten times the first day at SEGA, and I was like, “WTF? What did I get myself into?”
Michael: I’m sure all of us have experienced those uncomfortable talks around how do we get first position in the waterfall. I’m sure we’re all sick of them. I think header bidding will help with that conversation. But I still think it’s fairly resource intensive for our ad ops team to maintain header bidding solutions within our stack. The really good thing about header bidding, for us, is there are a lot of bad actors in our space, in terms of DSPs. So hopefully header bidding will weed out all the bad actors where they’re just looking at our inventory, not really buying our inventory. Or if they’re buying it, it’s further down the waterfall. We’re hoping that header bidding will put a stop to that.
Christy: Michael, can you explain why you would use multiple header bidding solutions at the same time, and the use case there?
Michael: For us right now, it’s really in the test phase to see what’s going to work for us. That’s why we’re continuously testing some solutions. These solutions with some of the big partners that we work with that are coming on board with header bidding, we can’t really ignore it anymore. It’s ad tech, it changes all the time and we have to move forward with where the industry is going. We’ll see better yield at the top of our waterfall but we still envision that the waterfall approach will still be there for us, and programmatic further down the waterfall.
Adam: I don’t know who wins this, if there’s a winner. We all win, actually, if parallel bidding takes off. Everyone right now seems to be developing some form of header bidding solution.
The question is: What’s the heavy lifting going to be? If header bidding is as easy as an SDK update on the versioning, is going to be the right solution. We definitely want to test. We’re still testing three rewarded video providers after a year, and trying to drill down to who’s the best one.
I’m sure we’ll do some testing with header bidding, but I’m hoping that it becomes very apparent to us who is the right company, and I’m sure the answer will be different for other people. If Fyber’s your mediation stack, then Fyber FairBid probably would be a great way to go. There are also independents out there like Max Ads, but that’s a bolted-on solution and it’s going to be interesting to see how that works within a pre-existing mediation stack.
Christy: What piece of advice would you tell publishers who are evaluating header bidding solutions?
Brian: Transparency in the marketplace. Making sure that all the advertisers or networks or DSPs that are plugging into their solution are on an even playing field on both the buy side and the sell side. If it’s not fair, then advertisers will go somewhere else where they can get a better deal to buy inventory.
As publishers, we will figure out who is the most transparent, who has the lowest fees, on both sides of the coin. And publishers will follow the money.
Michael: I agree with transparency. We’ll now be able to understand our value of the user base, instead of guessing what our users are worth, which is really crucial for business.
Adam: One of the things you want to look for is how many people do they actually have integrated into their solution, because that’s a tough conversation for them. The demand side, not everyone wants to get in on it. AdColony actually is very advanced in that they actually seem to be in all of these solutions. They see the future. Some demand sources don’t. They’re worried a little bit about pricing—that they’re going to be paying more for impressions over time.
If desktop is anything to go by, it’ll normalize, but initially they may be paying more for that user, which is great news for publishers, not so good for them. But mediated ad networks will have to be in the fight if they want to continue because header bidding is going to take off, and they won’t be able to not be part of that.
There are three components to any header bidding solution. One is how many mediated SDKs do they have in there. And an SDK solution that’s able to do all the things you love now like rewarded video, but in real-time bidding. Secondary to that, and I know that Fyber FairBid has a tool to offer this, is other SDKs using historic information that can also bid in, but if it was like $9 CPMs they were bidding yesterday that would go in as their real-time pricing, even though it’s not real-time pricing. And then how big is their DSP marketplace.
So that’s what you want to be looking for to make sure it’s a robust solution.
Filippo: There is a more fundamental problem to all of this. Where I think we’re actually not all winners but all losers. I would say to everyone, go and talk to the IAB, and tell them to do OpenRTB on mobile. Because as long as we all measure things differently, be it for attribution, be it eCPMs, or the install is measured at the end of the video or in the middle of the video, and whatever other magic standard that each network or DSP has, then we’re not all playing on an equal level field. We’ve been harping on about transparency for years, but nobody seems to be trying to do anything about it. And it’s not really up to us or up to Fyber, or anyone else out there. It’s really up to a more independent body. Let’s try to come together as an industry and try to figure this out.
Brian: What publishers really want to get away from is when a buyer throws up like, say, a dollar on a bid and then the DSP takes 10 cents, the SSP takes another 10 cents, and the publisher ends up with 50. Where did that go? Where’d the rest go?
We’re just trying to figure out where those fees are going, what they are, and know more about that. Once all of the dust settles, we’ll know who comes out on top.
Adam: We should also be talking about technology implications of this and latency. If you’re doing real-time pricing on rewarded video, at what point is the provider going to cache that video? Or are they going to cache it, or is it going to take a hell of a long time, and so you get a winning bid but then the falloff rate is immense. So pay attention to the technical chops of the shop that you’re working with.
Christy: David, what is the buyers point-of-view on in-app header bidding?
David Pokress, AdColony: So they’re talking about the publishers’ side, I’m coming from the ad network side. And they talk about transparency, I talk about transparency. That’s probably my single greatest frustration with the existing waterfall setup is: I don’t know. And then I have to call and bug these guys and say, “What do I have to do to get my first position deal?” And I do that 14 times a week, and they get irritated. It takes a lot of staffing to do that. And the reality is: just tell me. I mean just tell me what I need to do, and I’ll figure out how to go do it.
AdColony sees in-app header bidding as an opportunity for transparency and equal opportunity. We deliver a lot of fill. My first impression could be worth $50, but because I’m being judged in an aggregate eCPM, it could bring that aggregate eCPM down to $12. AdColony could afford to bid $50 for that first impression if we were able to disaggregate all that stuff out. I think that’s going to be a big advantage so I’ll be able to compete. There’s no more deals for first position. Either I earn it or I don’t earn it.
We are bullish on this future. We think in-app header bidding is good for publishers. We think it’s good for us. Ad tech is incredibly complicated—almost too much so. And I think it goes to Brian’s question: Where does that money go? It goes down the ad tech rabbit hole, and that’s not always a good thing. The more transparency, the closer we are to each other, bidding will be better off.
Christy: Where does the CPM increase from in-app header bidding come from? Does it come from demand source diversification? Better yield management? What are your thoughts?
Brian: Right now, if AdColony is not your first look, you really don’t know where they need to come in at. But when we transition on over to this new model, they’ll be able to look at the total amount of bids that they placed and look at what the percentage is on what they won, and where they need to bid to get a higher win percentage. AdColony will automatically know versus actually having to email or talk to publishers all day long, and sling deals on the side—where that’s not as efficient. Hopefully buyers will know exactly where they stand, and what price points that they need to bid at to win that inventory that they’re looking for.
Adam: If I worked at a DSP, I would be incredibly frustrated by the waterfall because here I’ve got a $25 CPM, let’s say, for an automotive advertiser. SEGA doesn’t know that that person’s in the market for a car, but that DSP sure does. But the DSPs never get to see them. Because AdColony got it.
Filippo: It will, exactly as you said, be diversification of demand and will allow even non-performance, more brand-led, non-gaming oriented demand to come in and have a fair chance. But at the end of the day, if advertisers are still not getting quality, that will be a temporary 30-40% CPM lift and then we’re back to just performance ads from games. And hopefully it will also help diversifying in non-tier 1 geos that tend to be more difficult to monetize. But again, with a little bit more transparency, it would be clearer to understand how we can deliver more quality to those brands, or increase monetization to the non-tier 1 geos.
David: I don’t know where this is going. I don’t think anyone has enough experience yet to say how successful in-app header bidding is going to be, whether it’s a 30% or 40% CPM increase. But I will say, another reason why I like transparency, is I know what I’m bidding on. We make a lot of bad decisions day in and day out. I make a lot of super irrational decisions, like I need to beat AppLovin or Vungle. I’m going to bid really high just because I really want that inventory for my demand team. I make irrational decisions all day long.
In the new world, I’ll be able to make a lot more rational decisions. I can say, “What is this really worth? Is this the first impression of the day? Is this the first impression of their session?” I’ll be able to have so much more information at my disposal to say, “What is this really worth to me?” AdColony has a lot of brand demand, and we use that, and it’s great. It drives up our eCPMs. But you know what, brands don’t care about position as much as performance does. If I can buy the sixth or seventh position for pennies on the dollar and be making $25-$30—that works out for me, but that currently goes into the publisher’s pocket. So that kind of imbalance, right?
Just because we’re going to bid, we’re going to get more aggressive and it’s going to come from the networks. It’s interesting to think about where else you can increase your revenues, whether it’s getting more fill globally, in two or three territories or whatever it is. That competition that we have that comes from non-transparency, I think is real. I mean, I think we all do really bad things just to stay competitive because we don’t know. We trust what publishers tell us, not that you guys ever lie, but there’s nothing to stop you guys from saying, “You need to hit $25 eCPM,” and it’s really $15. And we jump. You need to be honest.
Adam: So just one counter. We don’t know where this is going, you’re right—but past is prologue. In the desktop world, we saw exactly what happened and there’s no reason it won’t happen here. There’s a reason why 80% of online publishers in the desktop world have a header bidding solution because they did see 20-40% CPM lift.
Michael: I think header bidding is going to help get the right ad to the right user as well, which is going to help drive up the revenue for both sides of the house. I think most of us are sick of seeing Machine Zone ads when we’re never going to convert.
Christy: AdColony was obviously one of the first traditional mobile ad networks to build the tech that’s necessary to buy programmatically. Why did AdColony decide to make that investment now?
David: It comes from a frustration with the current system and setup. We were pushing RTB hard for well over a year now. A lot of the initial friction was people were fine to do a RTB solution, but in our particular case, if we win, we want to fire our SDK. And that was a bit of friction with some of the existing partners that do that type of programmatic solution. They’re not set up to fire somebody else’s SDK upon winning. But we sell a premium ad product that requires our SDK. We use that for data collection as well. So we knew that we needed to start talking early and convincing people that in-app header bidding is a good idea to maximize revenue.
We bring a lot of demand, and we know that being judged on the aggregate is not necessarily a great way of doing anything. Any way to disaggregate the eCPMs is good for us, and in-app header bidding is the solution that delivers that.
Christy: Does in-app header bidding change the way UA teams approach their strategies?
David: I don’t really know yet. Obviously, our dominant method these days is CPI. In a bidding environment, we’re not doing a soft bid, we’re actually doing a real bid. I’m paying on a per impression. I may get zero revenue from that impression. It’s not eCPM, it’s CPM. It’s not an effective CPM. I’m actually saying, “I’m paying this.” That’s a big mental shift for a lot of people. We were already stack ranking our demand that way anyway, so it’s not too far of a stretch to start bidding.
David: But if the demand team is super excited about some supply, they start bidding CPM anyway. So we’re there, it’s just a matter of delivering against whatever the advertiser wants. And if they’re driving towards ROAS (return on ad spend), they’re driving towards a certain cost per install or a combination of all that stuff—it’s just about hitting those metrics. Again, greater flexibility of what we bid in, and with full transparency, we’ll do a better job delivering against our advertisers’ needs.
Filippo: If in-app header bidding delivers as promised, it will be a game-changer.
Basically, it might be based on what level of risk you can tolerate. It might be better to go to the DSP, and bid on CPMs instead of buying on CPIs directly with AdColony. So it could shift the relationships as well. Whether we have direct dealings with the ad network, or we go via the DSP. So we can all tap into and bid competitively on CPM. So that’s all to say: How is this going to work out, as David was saying, is still difficult to predict. But the fact that we will start moving away from just purely CPI and taking on more risk ourselves, and going to CPM, could be significant.
Part of the problem that we have is that we’re only willing to accept CPI for gaming in general. We’re just not willing to take the risk. And it could be interesting to see how it pans out with brands who don’t care about CPI.
Christy: How will mobile ad networks manage the risk element?
David: Every impression that runs that doesn’t make money hurts everybody. It keeps eCPMs down, and it’s inefficient for AdColony. We’re always “managing the risk,” I suppose, if you want to think about it that way. It’s just about being efficient—right ad to the right person at the right time. It doesn’t get much more cliche than that. And with transparency and the right information, you’ll do a much better job getting the right ad to the right person at the right time, for the price you’re willing to pay. And that’s only an improvement over what we currently have today.
Adam: Don’t think that in-app header bidding is a silver bullet. Things won’t change overnight. The waterfall is still going to unfortunately be there for, I imagine, at least two years. And maybe it never goes away fully. In-app header bidding is another weapon for you to use to drive up revenue incrementally.
Brian: David also touched on a really specific point in what publishers should be looking for when they’re choosing a technology like this. AdColony’s SDK needs to be included, but there are also a lot of vendors out there creating these technologies in where it’s SDK-less or just one SDK. A publisher might be on S2S, but if AdColony isn’t buying through an S2S connection, then that publisher might lose all that demand from AdColony. You need to make sure that whatever in-app header bidding provider that you’re working with, allows those SDK integrations to still be there. I don’t think, as a publisher, we’re going to move away from SDKs anytime soon either.
Christy: Are there in-app header bidding solutions that you’ve encountered that don’t include the SDK element? Do you allocate ad budgets differently if it’s doesn’t?
David: No. The conversation has been about how do we make it so that the SDKs are part of the solution. Again, when we first started these conversations 12-18 months ago, that was like a push.
I feel like everybody is embracing in-app header bidding. But I get the challenge of the SDK and, as long as it’s delivering value, there’s a reason to include the SDKs. If it reaches a day where the SDK doesn’t deliver value, then publishers should all kick it to the curb. But it’s our job to make sure that we’re delivering the highest revenue to the publishers and, right now, we can do that via an SDK.
Adam: This is why I don’t want to call it header bidding though. One, there’s no header, so it’s stupid to call it that. The second thing is, it’s distinct. Being new to the mobile ad tech industry, I realized just how different it is from the desktop and mobile web space. So having the SDKs as part of this is a very important thing, and it makes it unique as a sort of real-time bidding solution.
Christy: What is a unified auction and how is it different from in-app header bidding?
Brian: To be a true unified auction, that platform, or whatever header bidding solution you’re using, should be agnostic to all types of demand. If you’re using Fyber, you should allow MoPub to bid in, and vice versa. The fees, again, should be transparent, fair, and even across all DSPs or ad networks, regardless of how much spend that are flowing through that bidder’s pipes. And on the pub side, the fees should also be transparent to make this very fair and even for both sides of the coin.
Filippo: The difference is just labeling, in my view. It’s just changing the sticker on the box.
Michael: It’s about allowing everyone to compete on an even playing field. With these hybrid solutions, some of them are still using historical CPMs with header bidding on top, and it’s just not a fair competition. So probably leaving money on the table when it’s that hybrid solution and not a true competitive auction.
Adam: I just want someone to figure this out for me so I can make more money for SEGA.