Beginner's guide: From in-app header bidding to unified auctions
by Scott Reyburn, Director of Content Marketing
It’s rare to see evolutionary change in action because it occurs gradually and incrementally over time. Sometimes, however, those changes can be seen ripping across the landscape like a lightning storm. Such was the case when the digital advertising industry witnessed evolution before its very eyes. In mid-2015, header bidding came to be, shocking the desktop advertising world. Two years later, the header bidding storm had nearly washed away the inefficient—yet dominant—waterfall model, achieving an over 70% adoption rate among the leading North American desktop publishers, while at the same time growing CPMs by the double digits.
Desktop publishers were tired of Google’s DoubleClick for Publishers (DFP) ad server giving their own ad exchange (commonly referred to as AdX) an unfair advantage over other demand sources in the waterfall, thereby artificially limiting the revenue potential of each ad impression. So the collective of publishers in the digital ad industry stood up and said, “No more.” Publishers’ answer to this bias: header bidding. As the name implies, header bidding took to the header (as in the head element of a website) to circumvent DFP, where an auction would be conducted in real time among preferred ad exchanges—particularly before calling the DFP ad server.
Today, mobile app advertising is undergoing an evolutionary transformation of its own, though it’s an entirely different beast compared to its desktop and mobile web counterparts. Mobile app advertising requires more than just a new auction to improve fairness and efficiency, it requires a complete overhaul—all the way down to its ad serving genetic blueprint.
In this guide, we explain the genetic underpinnings of mobile app advertising’s shift from the sequential waterfall model to a parallel, unified auction, so you can prepare your business for the future of mobile ad buying and selling.
What is in-app header bidding?
Some words just cause a stir. Take, for example, “in-app header bidding”—it’s a sit-up-in-your-seat-straighter word right now in the mobile advertising space, and for good reason.
Today, however, the term “in-app header bidding” is being used as a boogeyman by tin-foil-hat-wearing cynics to scare away mobile app publishers and advertisers from believing that it will bring fairness, transparency, and efficiency to the mobile advertising industry. Those same skeptics use the literal translation of “in-app header bidding” as a red herring, quickly pointing out that there’s no head element in a mobile app—and therefore in-app header bidding will be no more than a “hack” like it was on the web. Literally speaking, yes, there is no “header” in a mobile app.
But let’s, for a moment, wax philosophical about the term “in-app header bidding.” In one memorable scene in the movie, The Matrix, the aptly named Spoon Boy, who was in the middle of bending spoons with his mind, shared with Neo one invaluable lesson: “There is no spoon.” Long story short, Neo was being taught that he can’t focus on changing things, he has to focus on changing himself. So don’t take the word “header” in “in-app header bidding” at face value. Rather, focus on the underlying concepts in-app header bidding was built on—fairness, transparency, and efficiency.
There Is No Spoon / Via youtube.com
With that said, let’s set the record straight and define in-app header bidding once and for all. In-app header bidding (also referred to as parallel bidding or unified auction) is a programmatic selling technique that allows mobile app publishers to offer ad inventory to all demand sources at the same time. Viewed through the prism of media buyers, in-app header bidding is a programmatic buying technique that allows media buyers to compete fairly on every piece of ad inventory at the same, not just the small percentage of inventory they would see based on their rank in the waterfall.
Unified auction: What it is and what it isn't
The underlying ideas behind in-app header bidding—like many ideas—can be twisted and abused, as noted in the previous chapter. Despite the attacks on in-app header bidding by naysayers, the concepts it represents—that any and all types of demand sources can bid at the same time—can’t be destroyed. In-app header bidding just needs a more suitable name to shed its association with its web counterpart, header bidding. Parallel bidding, advanced bidding, and FairBidding, are just a few of the names that refer to a similar concept, which is essentially the concept of a unified auction.
A true unified auction is unified in every sense of the word. It’s more than just the intention to bring fairness, transparency, and efficiency to mobile advertising, a unified auction is an advertising trading mechanism built from the ground up for the unique, though fragmented, mobile app environment.
What is a unified auction?
A unified auction environment for mobile app publishers is a demand-agnostic auction facilitated in real time, wherein each ad impression is simultaneously offered and sold to the highest bidder from among any type of demand source, including client-side (i.e. SDK) mobile ad networks with (and without) RTB capabilities, server-side programmatic demand sources such as DSPs and ad exchanges, direct deals with advertisers, and cross promotion.
“To be a true unified auction, that platform, or whatever header bidding solution you’re using, should be agnostic to all types of demand,” said Brian Kealer, director of ad monetization at Glu, this past March at the 2018 Game Developers Conference.
In a real unified auction (above), like that of Fyber FairBid, each ad impression is simultaneously offered and sold to the highest bidder from among any type of demand source.
A true unified auction environment embodies these characteristics:
Inclusive of all demand: No demand source is left behind. A unified auction, includes any and all types of demand sources, particularly bringing together server-side programmatic demand sources and client-side (as in SDK-based) mobile ad networks into the same auction to compete simultaneously for every ad impression.
Draining the (waterfall) swamp: A unified auction doesn't pour ad requests downward, like that of a sequential waterfall model, in order from the demand source with the highest expected eCPM to the next highest, and so on. Rather, each ad request flows evenly and simultaneously across any and all types of demand sources—both server- and client-side sources included.
Real-time bidding: In a unified auction environment, all demand sources bid in real time. As a result, app publishers can see exactly how much each demand source is willing to pay for their inventory on a per-impression basis.
What doesn't constitute a unified auction environment?
It’s almost as important to explain what isn’t a unified auction as much as what is. Now that you know what a unified auction is, here are the characteristics of anything that isn’t a unified auction.
Lacks diversity: For many app publishers, mobile ad networks without RTB capabilities still make up the lion's share of ad revenue. Recently, some unified auction environments are being passed off as ones that only include RTB-capable demand sources. Another iteration of a unified auction environment doesn't include SDK mobile ad networks into the equation at all—RTB-capable or otherwise.
There's still a waterfall: Where the highest bidder from a first so-called "unified auction" is sent to a publisher waterfall in any sort of ad serving platform to compete again, where manipulation at the hands of the ad serving platform could still occur.
Sequential mediation: The current state of mobile app advertising affairs
To understand the importance of the mobile advertising industry’s evolutionary transformation to a unified auction environment, it’s necessary to have a basic understanding of why the sequential waterfall model used by mediation platforms today is unfair and inefficient for app publishers. We’ll spare you an in-depth description of a mediation platform’s inner workings (see this blog post for a detailed account).
What is mobile ad mediation?
A mobile ad mediation platform lets app publishers manage how they want to sell their ad inventory among demand sources—chiefly SDK-enabled mobile ad networks—through a method called “waterfalling.”
Waiting in line
Mobile mediation at its core lets app publishers do two main things with the goal of selling their ad inventory at the highest price through increased competition:
Manage how app publishers want to allocate their ad space among various mobile ad networks as efficiently as possible—choosing to manually rank ad networks based on preference, or automatically through a mediation platform's automated yield optimization technology.
Allow app publishers to directly integrate a single mediation platform SDK, which streamlines the development work in order to integrate multiple ad networks.
Without making any code changes to their mobile apps and games, publishers can rank their demand sources in descending order of preference and traditionally have chosen to do this based on the highest expected eCPM to lowest (i.e., the sequential waterfall model).
Being ranked first means that a specific ad network—usually the one that has historically paid the highest on average—is given preference over other ad networks and has the first chance to respond to an ad request. The ad network ranked second in line can act on a request only after the first ranked network cannot, or chooses not, to return an ad fill. And so it goes for demand sources ranked third, fourth, fifth, and so on—each wait in this figurative line for the request to reach their rank in the waterfall.
Ranking based on historical ad revenue data makes it difficult for advertisers to move up the waterfall—past eCPMs are not the best indicator of future performance. This means the same networks tend to get the first look, resulting in some other networks never competing for inventory—even if they are prepared to offer a higher price for a given impression.
Example of flawed waterfall ad serving
Let’s say that the current waterfall order—based on historical data—is as follows:
Network A: $10 expected eCPM
Network B: $9 expected eCPM
Network C: $7 expected eCPM
Now let’s say that Network A is willing to buy this impression at $11, but Network C, which just brought on a new campaign, would actually be willing to buy at $25.
Under a mediation’s waterfall model, Network A would be called on first, and as long as it respond “yes” to the ad request, it would win that impression and pay $11. Network C wouldn’t even be called, despite having a higher-paying campaign. Not only that, but Network A would remain ranked first, as its average eCPM now increased. So not only did Network C not get a chance to pay $25, but the fact that it has a $25 eCPM campaign doesn’t even better its rank in the waterfall.
Cutting in line
It may sound like app publishers are getting a fair price by giving preference to the best performer based on average historical price. However, because most mediation platforms are free, in order to make money the mediation platform may be naturally inclined to favor its own demand (via an in-house marketplace or ad network) over third-party demand sources. In effect, the mediation platform is cutting in line, the same way AdX did in DFP, which, as you may recall, was the initial reason that led to the birth of header bidding in the first place (see Chapter 1).
There are two common ways in which mediation platforms cut in line:
A mediation platform giving its in-house demand a first-look advantage for ad inventory. The more ad impressions that come from a mediation's own demand, the more app publishers believe that it must be the demand source generating them the most ad revenue.
If a mediation platform's in-house demand derives from a programmatic RTB marketplace, then that platform can potentially have a last-look advantage over other demand sources (particularly mobile ad networks) in a waterfall setup, allowing it to win more ad impressions than it should. A mediation platform has this advantage because it can submit bids to an auction in real time, while, in comparison, third-party demand sources can't respond with a bid in real time at all—leaving its average eCPM to stand in as a pseudo bid in the so-called auction. Under these conditions, a mediation platform can "look" at each auction and make a last-minute decision to monetize an ad impression or not.
Today, most app publishers will say that using a mobile ad mediation platform is a no-brainer. But asked if they would use a mobile mediation platform not based on the waterfall model, publishers may ask themselves if there’s a parallel way to sell ads programmatically. The answer is “Yes.” The unified auction is the evolutionary shift away from the waterfall approach.
The benefits of a unified auction
Now that you have a better understanding of the unified auction’s evolutionary impact on the mobile advertising industry, it’s easier to understand its resulting benefits. Here are the three key benefits of a unified auction environment (in no particular order), all of which contribute to more ad revenue for publishers, as well as a fully transparent auction setting for buyers—free from in-house bias.
1. Understand the market value of ad inventory
Mobile ad mediation is an inherently flawed system because there’s some guessing involved, as outlined in Chapter 4. In a unified auction environment, mobile app publishers can see the actual bid price from any type of demand source that is willing to compete for an ad impression. Contrast that to the mobile advertising industry of today, where demand sources compete for ad inventory based on average eCPMs, which is a clear condition of imperfect competition.
The improved market conditions created by a unified auction, where any and all types of demand—including server-side and client-side—can compete at the same time and place a bid for each impression in the ad response, will increase yield for publishers. And as another result of the improved market conditions, app publishers can understand and grasp the current market value of their mobile ad inventory—thus being able to set more accurate price floors.
2. Demand-source-agnostic auctioning
No longer does a particular type of demand source get any sort of advantage over another. In a unified auction, any and all server- and client-side demand sources are treated as equal, including client-side mobile ad networks with real-time bidding capabilities, server-side DSPs (where many media buyers manage brand campaigns), direct sold buyers, and client-side mobile ad networks that have yet to make the transition to RTB.
3. Improved efficiency
Managing a waterfall manually isn’t a set-and-forget-it exercise for app publishers that consider mobile advertising a crucial source of revenue for their business. Yet many leading app publishers, including the likes of Glu, are trying to get away from the time-consuming waterfall management process. “That’s what we’re trying to get away from,” said Glu’s Kealer at the 2018 Game Developers Conference.
An important message for ad ops professionals about unified auctions: reorder no more. In a unified auction environment, ad ops teams will be able to collect and analyze actual bid request data, no longer having to rely on historical ad performance data. With this powerful data at their fingertips, ad monetization teams can better group, price, and sell their ad inventory.
AdColony's David Pokress on supporting the unified auction in mobile apps
David Pokress has worked with some of the most iconic brands over the past 15 years in the interactive entertainment industry including Call of Duty, Skylanders, DOOM, Tony Hawk, Marvel, James Bond, and Transformers. In his current role as executive vice president of global publishing at AdColony, Pokress works with mobile app publishers to help grow their business and work with AdColony more effectively.
David Pokress, executive vice president of global publishing at AdColony
AdColony, a global company with 500 employees in 12 offices worldwide, started as a rewarded video ad network and is now one of the industry’s leading mobile video monetization platforms. In line with their mission to help app publishers maximize their ad monetization potential, AdColony teamed up with Fyber as a mobile ad network launch partner for Fyber’s unified auction technology, Fyber FairBid.
In this exclusive Fyber interview, Pokress sheds light on unified auctions from the perspective of the buy side, namely mobile ad networks. Here, he explains why the evolutionary shift to unified auctions in the mobile advertising industry is a “huge opportunity” for both app publishers and mobile ad networks. David also shares the key benefits that mobile ad network SDKs will continue to provide in a unified auction environment.
A transcript of our interview follows, lightly edited for length and clarity.
Based on your years of experience at AdColony, can you paint a picture of the current state of the mobile advertising industry?
David: The current state of mobile advertising is strong, but it’s also very competitive. As a concept, it’s more accepted today than it has ever been. When I started out, we had to convince people that advertising was OK at all and that integrating an SDK as part of that process was OK as well.
Mobile advertising today is a fixture—it’s an established, viable way to earning revenue in mobile. It doesn’t impact retention or engagement, and it allows people to make more money from their users. The [mobile advertising] industry has matured from one where we were trying to convince people that it’s a good idea, to where it’s accepted and embraced wholeheartedly—with some people making all of their revenue through advertising. There’s just a lot more opportunity for the partners to work with the publishers than there ever has been.
Now, over the past few months, there has been an evolutionary shift in the mobile advertising landscape away from the waterfall model and toward RTB. Why is AdColony dedicated to making a real-time, unified auction in the mobile app environment a reality for the industry?
We’ve been very supportive of this idea for quite a while. In fact, it felt like we were pushing this idea long before many of our mediation partners embraced it. From our perspective, the ability to bid for every impression is a huge opportunity for both publishers and networks. Previously, with basic waterfalls, you were essentially being judged on your expected average, and you were being slotted into a waterfall position based on what your average does.
We have a lot of demand, and that involves trying to drive total revenue for our publishing partners—which means we have the highest of the high and more moderate eCPMs as well, depending on the advertiser and the campaign. The traditional waterfall gets influenced by both extremes, which can create a feedback loop that does no one any favors.
The adoption of a real-time, unified auction and this ability to actually compete for every impression is something we’ve been waiting for and hoping would happen. It’s how we all can compete more effectively for the supply.
"The adoption of a real-time, unified auction and this ability to actually compete for every impression is something we’ve been waiting for and hoping would happen."
In the mobile programmatic RTB environment, AdColony would be submitting committed bids for the first time. What steps has AdColony taken to prepare itself for bidding in real time?
Our technology has always assigned a value to supply. We were already converting CPI into CPM, in order to better serve our partners by maximizing value. So really, it’s not a tremendous change for us, quite frankly, because we were ahead of the curve a bit and already thinking that way.
What technology has changed under the hood of AdColony to support bidding in real time?
AdColony started out with a singular focus on gaming, a lot of which was performance-based user acquisition, and we’ve been working on the brand advertising side as well for some time. From 2014, AdColony was actually a part of a bigger organization that operated outside of the AdColony SDK. We were part of Opera Mediaworks, which had been operating in the programmatic environment for a long time. The AdColony you see today is the merging of those businesses and expertise.
The team here has a lot of knowledge and real-world experience operating in a programmatic environment. For us, there’s never been a huge mental or technical shift required because we were already operating in these environments—just not in the app environment. And that really was driven by the mobile app ecosystem’s reliance on ad networks and their SDKs.
The ad networks deliver a massive amount of value. For the programmatic solutions that people are used to dealing with, if you win, you stream the ad, whereas in the mobile app environment, in order to maximize value, you really need to call an SDK. That was really the biggest change that needed to happen within the industry: The recognition that, in order for programmatic to be successful with mobile apps, you had to work with ad networks, and that meant calling out to the SDK from the winning ad network in order to deliver the ad.
Mobile ad networks have historically been the leading demand source for app publishers. In order to access this distinct demand, publishers have had to integrate an SDK into their apps. In a real-time, unified auction environment, can you describe the key benefits that an SDK integration will continue to provide?
The SDK allows us to deliver the maximum eCPM. We have higher quality ad units that users prefer interacting with that are driven by the power of the SDK, therefore we can be more effective for our advertisers, which means higher CPMs, which means better results for the publishers.
How does the change to programmatic RTB affect your advertisers directly? As an example, what changes for an advertiser running a CPI campaign with a campaign goal of driving installs, but at the end of the day, that advertiser will compete in a CPM-based auction?
Nothing really changes. We work with advertisers on a CPM basis, we work with advertisers on a CPI basis, just like on the publishing side where our job is to support the publisher, and be flexible and operate in a way that makes sense for them. The same thing with the advertiser. If they want to operate in CPM, we can do CPM. If they want to operate in CPI, we’ll operate in CPI. We’ll make the translation into CPM in the bids themselves—that’s our responsibility and part of the technology that we deliver.
What’s the split between brand advertisers and performance advertisers on the AdColony network?
Brand and performance are split pretty evenly on our network, it’s about 50-50.
How will a unified auction environment impact brand advertiser campaigns?
Overall, the CPMs of the brand campaigns are relatively strong, so I don’t think they’ll have a challenge, necessarily. From our perspective, we’ll have more flexibility to decide where to put that brand ad. For the most part, performance advertising tends to be the most effective early on in a session with a user. So we can choose to bid high for performance campaigns earlier in a session, and then can afford to bid a little lower for brand campaigns later in the session.
Brand campaigns do have performance metrics, so obviously we need to be mindful of those. For the most part though, [a unified auction] will give us greater flexibility about when to bid and where, and how to operate within a specific publisher as opposed to being stuck within a narrow band.
"For the most part though, [a unified auction] will give us greater flexibility about when to bid and where, and how to operate within a specific publisher as opposed to being stuck within a narrow band."
Can you explain why it’s such a game changer for AdColony to move from competing for inventory in an opaque waterfall setup to a real-time, unified auction environment, where you’ll have unprecedented transparency and visibility into each ad buying decision?
You hit it on the head with transparency. Aside from being able to bid for every impression, having a complete view of what it takes to win each impression is a huge opportunity for ad networks. In the traditional, more opaque environment, it was very challenging to figure out exactly where you needed to operate in order to move up the waterfall. It’s dynamic, so it can be difficult to keep up with anyway, but you’re really relying on secondhand information to know what it takes to get to first position if you’re in second position. Unified auctions remove a lot of that guesswork.
It really will be a huge opportunity for us just to understand individual mobile app environments better, and discover what it takes in order to be successful in each one.
In a unified auction environment, how do you see the relationship between the ad network and publisher changing? Will dealmaking still play a role?
I think many of those functions will still exist in the new world. There’s still a lot of control that a publisher has, even with advanced mediation—and they can always choose how to implement it. At the bottom end, our job is to work to optimize our offering to maximize the publishers’ revenue. So, if it’s done through deals that maximize eCPM and maximize waterfall position, that ultimately provides more revenue for the publisher, and we’ve accomplished our goal. That’s one way of optimizing.
Or, it’s optimizing our bidding to win more auctions and maximize revenue for the publisher that way. It’s still working to optimize or to maximize revenue. It’s just how we do it that will be slightly different.
I’m not sure deals are going to go away 100%. There will always be that opportunity where it makes sense for the publisher to strike deals. I think that opportunity will still exist, and we’ll still have to look at that going forward.
Do you have a final point that you want to raise about the unified auction and header bidding for mobile apps for our readers?
"The ad tech universe outside of in-app mobile has continued to evolve, and header bidding was one step in that evolution."
I think everyone is just going to have to do trial and error in testing and figure out what are the best implementations. Certainly, the ad tech universe outside of in-app mobile has continued to evolve, and header bidding was one step in that evolution, but people are always looking for different ways to maximize revenue. The widespread adoption of in-app header bidding for mobile ads is an important step for our industry, but it certainly won’t be the last.
The future of bidding and unified auctions
Bias, inefficiency, and lack of transparency are what the mobile advertising industry has come to know and tolerate from the waterfall model. However, the world of waterfalling is crumbling down before the industry’s very eyes. The unified auction is rising from the waterfall model’s ashes—a new, real-time auction environment free of unfairness, inequality, and non-transparent practices.
"The unified auction will continue to evolve over the coming months and years, from more than just the unification of the auction environment for buyers and app publishers."
The unified auction will continue to evolve over the coming months and years, from more than just the unification of the auction environment for buyers and app publishers. In the future, tools and dashboards will be unified. Reporting will be unified. And ad standards will be unified. This future isn’t too far away, and we’re excited to be a taking a step toward a more fair, transparent, and efficient landscape for buyers and publishers alike.
Alina Lacey-Varona contributed to this guide.
Design by Kelly Howe / Kyuhee Jo.