The Buy-Side Guide To Header Bidding

Buyers-Header-BiddingAs publishers embraced header bidding, many programmatic buyers observed the trend from the sidelines. Since header bidding boosts yield for publishers by increasing competition, it follows that buyers may be paying more.

But buyers wouldn’t necessarily know if that’s the case, since SSPs don’t note whether the impression comes through a header or not.

If header bidding drives up prices for buyers, it’s possible that it’s balanced out through improved reach and quality. Header bidding also increases programmatic supply, making it easier to find and bid on consumers.

But buyers are ambivalent about header bidding, which increases complexity, technology costs and – because it offers up the same impression to multiple exchanges at the same time – creates the potential to bid against oneself.

As a result, some buyers are changing how they bid and connect with publishers, whereas others are planning to join buying platforms like Criteo, Amazon and the DSP AudienceScience and create their own publisher networks to ensure priority access.

AdExchanger spoke to agencies, buy-side tech and consultants in order to figure out how they’re reacting to header bidding.

Are CPMs Actually Increasing?

“Get ready to pay more,” said Jay Friedman, partner and COO of the Goodway Group, which advises agencies on programmatic.

While some buyers see price increases they attribute to header bidding, not all of them mind. Some said increased prices came with better quality or improved acquisition metrics.

MediaMath sees higher prices when bids pass through header bidding, at least from the exchanges that share that information.

“It’s form of privilege, and privilege comes at a price,” said MediaMath VP Tanuj Joshi.

Others on the buy side, however, say header bidding artificially increases auction prices, and they’ve adjusted their bidding strategies because of it.

Criteo – once the sole header bidding partner for thousands of publishers – attributes some price increases to the rise of header bidding, but it decreased its bids because it didn’t see better value.

“The more non-second-price the auction is, the more you bid down,” said Patrick Wyatt, SVP of product management at Criteo.

Buyers who only bid what they’re willing to pay aren’t concerned about the difference between what they bid and where the auction clears.

“Intellectually, it doesn’t bother me that one [impression] clears at $5, one at $7,” said Eric Bosco, CEO of ChoiceStream. “My value is my value.”

Buyers also don’t mind paying more if it allows democratized access to quality inventory.

“We see header bidding as a shift to make more auctions more fair,” said Rocket Fuel CTO Mark Torrance. “It’s better for the publishers, and also better for the demand side who would not have access to that inventory.”

Unfortunately, the same pricing mechanics that lead buyers to pay more could also lead them to bid against themselves.

Imagine that a buyer submits a $5 bid to four different exchanges for the same impression, which clears at $2, $3, $3.50 and $4. The publisher would select the $4 bid to show the impression. But if the buyer hadn’t submitted that $4 bid, it might have ended up at the $3.50 clearing price.

The other way a buyer might bid against itself is when a header partner passes that $4 bid into the ad server. Google Ad Exchange has the final opportunity to beat the bid via enhanced dynamic allocation. The same buyer may pay $4.01 or higher for the impression it bid on earlier.

GroupM Connect is not concerned about this.

“If you would have asked me five years ago, I would have said that [bidding against yourself] is a major risk,” said Joe Kowan, global head of programmatic at GroupM Connect.

But research from GroupM’s Exchange Lab “proved that taking precautions to mitigate [bidding against yourself] would hinder campaign performance multitudes higher than bidding against yourself,” Kowan said. “It’s rarely even worth considering unless you are a massive advertiser.”

Tweaking bidding strategies and publisher connections to accommodate header bidding takes resources from other priorities.

“If I’m an agency trading desk and now have to develop a whole new set of intelligence around bidding algorithms, how am I making those fees?” Friedman said. “This is probably the biggest leap in complication I’ve seen in a while in terms of being able to truly understand what you are bidding on, why and when.”

Will Header Bidding Change How Buyers Connect To Sellers?

Some buyers want to use header bidding tech to connect to a select group of publishers.

While directly connecting to publishers isn’t new (Criteo and Amazon connect directly with thousands of publishers, which reduces tech fees and allows funds to be reallocated to publisher acquisition teams), some direct connections let DSPs cut out SSPs.

AudienceScience, for example, began pursuing direct header integrations for clients like P&G five years ago. The setup allows advertisers to “buy from the premium publishers, buy using their data, and with universal frequency capping,” said Michael Greene, VP of product strategy at AudienceScience.

MediaMath is also working on a solution that leverages technology from its Akamai acquisition, which solves some of header bidding’s infrastructure and security issues, said the company’s VP of product planning, Eric Picard.

“We want our demand to get as close to supply as possible, with as few hops or intermediaries in the middle, and to deliver ads, not ad tech,” MediaMath’s Joshi added.

Buyers who believe in the open exchange but want to avoid second-pricing themselves are culling their suppliers. While DSPs used to establish connections to as many supply partners as possible, they may soon start buying through a smaller number of exchanges.

Because the same impression can come through exchange channels five or 10 times, leading to price increases, DataXu is revisiting how it accesses publishers, said company co-founder and CTO Bill Simmons.

“DSPs like DataXu need to strike preferred partnerships with preferred exchanges and require them to send publisher inventory through the exchange only once – while blacklisting it from other exchanges,” Simmons said. That way they can avoid the waste of bidding against themselves.

Similarly, Criteo is going to withdraw demand from SSPs because of header bidding. “We will have more incentive to run direct deals,” Wyatt said.

GroupM Connect, which already focuses on private marketplaces and works closely with publishers, hasn’t changed its strategy, Kowan said. When it sees strong performance from a header tag in an open exchange, it uses that as a conversation to start a direct integration.

Kowan is not surprised to see more DSPs go GroupM Connect’s route and focus on curating supply instead of fishing in the open exchanges.

“DSPs [are] starting to pivot and go the direction we have and build direct relationships with the publisher. They realized that these open exchanges are not what they are cracked up to be,” Kowan said.

Agencies can curate header bidding supply more manually, too, on a campaign-by-campaign basis. The independent agency PMG pays close attention to which publishers have header bidding integrations, because these setups improve reach.

When it assembles a media plan, PMG will consult with its DSP and target spend toward publishers running header bidding. “We are always trying to figure out a way to buy qualified, good inventory,” explained media operations director Cristina Freeman. Header bidding simplifies that process.

Plus, the company doesn’t have to deal with the hassle of setting up private marketplaces. That said, it’s receptive to publishers who reach out and want to establish a direct relationship if a PMG client is buying a lot of inventory through the header.

But not everyone likes the idea of buyers focusing on differentiated supply. ChoiceStream staffs a small publisher acquisition team to make sure it can access enough of quality publishers’ inventory programmatically. But it doesn’t like it.

“It’s disheartening, because it’s back to unautomated,” Bosco said. Establishing direct relationships with publishers creates “essentially an old-school ad network.” He prefers focusing on tech over supply. “Fundamentally, you can differentiate yourself on inventory or data. We picked the data front.”

DSPs moving into the territory of SSPs worries Friedman. Although some could be lured by access to unique supply, publishers technologically need to limit total partnerships.

“I prefer a hard line between the buy side and the sell side,” Friedman said. “I would not want my DSP having a header tag and having that direct relationship with the publisher. That goes over the line for a lot of folks.”

Whatever the outcome, header bidding is ushering in a stage of transition for buyers and sellers as they figure out what to do as their access to each other improves, but via another imperfect method.

 

 

6 Comments

  1. The buy side response to header bidding is going to be interesting; I still think buyers are formulating strategy, but they could take a number of drastic actions that impact the benefits of header bidding. What's clear to me is that header bidding has exposed the benefits buyers enjoy from an illiquid auction - that is, depressed prices. Time will tell if buyers or exchanges will work to remove some of the liquidity from the market as a response or if they'll take another approach.

    Sarah, on your point on buyers bidding against themselves, I'd offer a different view - logically you aren't bidding against yourself unless you are both the highest and second highest bid in an auction and you alone affect the price reduction on your winning bid. In fact I'd say the sailors second price auction dynamics actually protect buyers from that. It's true that buyers can pay more by bidding through multiple sources, but it's the other demand in those sources pushing the price up. The exception you have right is AdX, which can use a winning price from a header bidding parter as a floor. Then a single buyer could conceivably floor their own bid. I'm not sure why buyers wouldn't strictly prefer other SSPs for header bidding domains for this reason.

    The point from Criteo is interesting - if prices are rising it is because there is more demand. Bidding less might allow them to win the same volume but I can't see how it would be the same kind of inventory quality. If their price stays constant they're likely just moving down the demand curve - it's hard to see how performance won't also suffer in the long run.

    Reply
    • I think header bidding auctions are always first-price, so it's not the other bids from the same buyer through other exchanges that are setting the final price. What's really happening in the example is the buyer's final price is determined by the highest second-price across all the exchanges (i.e. the buyer bid at $4 because one exchange had another ad that bid $4 in the exchange's internal second-price auction). If the buyer hadn't gone through that exchange, the impression would likely have gone to the ad that came second in that auction. The buyer would never have got the impression for $2 or $3 - that's the whole point of the increased liquidity.

      Reply
  2. Header bidding is simply access to inventory and DSP and Agencies have always had access to inventory. The issue is this is better quality inventory and DSPs and the buy side in general have built their businesses and technology on acquiring inventory for as low a price as possible - namely remnant exchange inventory on AdX. Most of the DSPs don't have the CPMs or more accurately the demand necessary at those CPMs to be interesting partners to publishers. Certainly they do not have the technology necessary to bid and decision at the speed that is necessary to not impact latency for publishers. Lastly they can't find their audiences in mobile web which is the most interesting and most important part of header bidding. I'm not sure if Header bidding killed DSPs or this is a natural evolution but at this point the buy side guide to header bidding is business with their own proprietary first-party demand and one aggregator of all the DSPs.

    Reply
  3. Victoria

    'Because the same impression can come through exchange channels five or 10 times, leading to price increases, DataXu is revisiting how it accesses publishers'
    Having an impression come through an exchange multiple times is just as likely to depress prices for publishers - in general if it isn't filled the first time it is far less likely to be filled subsequently, not only because of simple supply and demand rules (artificially inflated supply) but also because the further down the waterfall an impression gets, the likely worse its viewability, load latency etc. Header bidding isn't that different to a waterfall (many publishers are already working with multiple SSPs), except that all demand sources are given the same opportunity to bid and win. This gives buyers better access to the valuable inventory. Poor inventory will still perform poorly.

    Reply
  4. Excellent perspective on an often overlooked side of header bidding.

    The #1 point of confusion that I see among buyers is the belief that header bidding unlocks incremental inventory access. For the great majority of header-enabled publishers, this just isn't the case. Programmatic buyers already had access to 100% of publisher inventory via AdX Dynamic Allocation. Those same impressions are now just being auctioned multiple times, create a false appearance of incremental supply.

    More on this here:
    http://jouncemedia.com/blog/2016/8/16/no-header-bidding-does-not-unlock-new-inventory

    Reply
  5. Re: buyers biding against themselves

    While true, this is not specific to HB. We encourage PulsePoint's buyers to submit multiple bids with each bid response. And yes, sometimes they do bid against themselves, but our test showed it is still worth it for buyers to submit multiple bids as they win more,

    some logic shall apply to HB

    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>