Header Bidding

Header Bidding Adoption: Three Years In Review

Pinterest LinkedIn Tumblr

Header bidding is here to stay. According to Adzerk’s HBIX tracker, its adoption is now at 80% and likely to trend towards 85%+. Very few ad technologies can claim such widespread usage.

Nobody knows where header bidding will go in the coming years, but looking at how it’s grown over the last couple of years could provide some insights. To that end, we’ve compiled historical data from the HBIX tracker. The tool measures header bidding adoption as well as client-side bidder and wrapper usage across the top 1K US sites.

Looking at the data collected over a span of three years, some questions came to our mind:

  • How has the header bidding market changed from 2017 to 2020?
  • What bidders gained the most?
  • And what does the future of header bidding hold?

We have addressed those questions below, looking specifically at header bidding usage across the top 1K US publishers (or sites that show ads).

Table of Content:

1. Header Bidding Adoption

In the tracker’s first report from the summer of 2017, we found that 70% of the top 1K US publishers (based on Alexa rankings) ran header bidding. This is an impressive stat, considering that the first Github release for Prebid.js was the summer of 2015, just two years before.

By 2020, that number had risen to 80%, a modest 14% growth over three years.

header bidding adoption

What does this data mean?

While we expected growth to be more, this is likely due to the high adoption rates header bidding already had in 2017. These numbers, for instance, roughly follow Geoff Moore’s concept of the Technology Adoption Life Cycle – in which approximately 16% of the population can be expected to adopt new tech only well after others have, if at all.

The data also highlights that header bidding is here to stay. Even if growth was relatively limited over the years, 80% adoption is impressive for a technology that is only five years old.

2. Unique Bidders or Adapters Found

The HBIX tracks adapters or bidders that are actively being pinged by a client-side wrapper upon page load. The header bidding endpoints – OpenX or Rubicon Project – make distinctive calls for header bidding and non-HB process, making it easy to differentiate.

As header bidding has grown, so has the number of found bidders.

Summer 201754 unique bidders
2020103 bidders
found bidders

This stat doesn’t tell the full story about how many heading bidders exist, as seen by the number of bidders listed as Prebid modules. These numbers will naturally be higher than above, as not all vendors who have bidders are actively used by the top 1K sites.

2017103 header bidding bidders
2020272 bidders

So, over three years, we saw nearly a doubling of unique adapters being actively used and a 160% increase in ad networks/exchanges building header bidding functionality into their product.

What does this data mean?

It’s clear that during this period publishers were giving the market a clear ultimatum: “We no longer want to rely on clunky waterfalls. Either offer us a header bidding endpoint or we’ll drop you.”

It should also be noted that the tracker looks for client-side ad calls; a customer working with an S2S wrapper like Amazon’s Transparent Ad Marketplace may be pinging more bidders server-side, but we have no visibility into that blackbox.

3. Average Adapters Per Header Bidding Site

Here, our hypothesis was that the growth of header bidding endpoints and the desire to maximize revenue would lead to steady growth in the number of adapters each header bidding site used.

Our guess proved true:

2017Sites that used header bidding actively pinged an average of 4.9 adapters
20207.2 adapters
average number of adapters per site

That’s 47% growth and reflects a core goal behind header bidding: if you can ping more demand sources simultaneously, your CPMs are bound to increase.

What does this data mean?

Publishers are always confused when it comes to the ideal number of bidders to add to their header auction. While a large number can increase page latency, a low number would cause a revenue decrease. In 2020, the average header bidder pinged seven partners. So if you’ve been steadily adding bidders over the years, you’re not alone. This doesn’t mean this is the ideal number for all publishers, though we suggest running tests to find what works for you.

One metric we track is what codebases are being used to build header bidding wrappers. In 2017, the open-source Prebid.js comprised about 50% of all wrappers found. But would another open-source competitor, such as Pubfood, enter the market and steal its market share?

The answer was a resounding ‘no’.

20172020
Pubfood1.6%0.45%
Proprietary48%28.4%
Prebid50.4%71.1%
2020 codebase breakdown

What does this data mean?

These stats tell a clear story: as time went on, sites and third-party vendors migrated away from custom header bidding solutions and coalesced around Prebid.js.

Such movement is great for the industry: publishers and bidders alike don’t have to worry about reinventing the wheel or spending time building multiple endpoints. Since Prebid.org is an open-source, independent, and transparent software, nobody has to worry about hidden fees, algorithms that may favor some partners over others, etc.

Prebid’s huge advantage here also indicates that it’ll be difficult, if not impossible, for any other client-side solution to eat into its market share.

Who won the client-side wrapper war?

While about 37% of sites in 2017 built their own header bidding wrapper (either with Prebid or a custom codebase), 63% outsourced the tech to a third-party vendor who managed the integrations on the client’s behalf.

By 2020, that percentage had dropped to 26% in-house, with 74% of all wrappers being outsourced.

Impressively, in 2020, we found a total of 89 different managed solutions being used, a fragmented market to say the least. Indeed, only one vendor, Index Exchange had more than 4% market share.

2017 most commonly found client-side wrappers:

WrapperPercentage share
Index Exchange44% of all third-party wrappers found
Prebid Enterprise (Xandr/AppNexus)8%
Curse Digital Media5.7%
Sortable5.2%
Proper Media3.9%
Freestar2.3%
2017 most found wrappers

2020 most commonly-found client-side wrappers:

WrapperPercentage share
Index Exchange30.5% of all third-party wrappers found
33Across4%
Sovrn3.7%
Freestar3.4%
Rubicon Project2.9%
Prebid Enterprise (Xandr/AppNexus)2.1%
2020 most found wrappers

What does this data mean?

The winner here is clearly Index Exchange. While they did decrease from 44% of wrappers found in 2017 to 31% in 2020, they still were the most-found third-party wrapper by a huge delta in both years.

The fact we found 89 unique wrappers also highlights the seismic shift that ad tech solutions have taken to stay abreast of current technologies. Whether one is an ad network, an ad exchange, a widget with ads, a managed monetization service, and so on, header bidding has created monetization opportunities for everyone.

5. Winner of the Bidder War

In 2017 it was no surprise that AppNexus was the #1 most found client-side adapter in header bidding wrappers, given that AppNexus pioneered the creation of the open-source Prebid.js. (AppNexus has since been rebranded to Xandr). At the time, 62% of header bidding sites were pinging AppNexus, with Index Exchange at #2 with 56%.

Bidder usage was certainly top-heavy too: only 13 ad exchanges were found in more than 10% of wrappers.

In 2020 that number improved quite a bit, with 19 exchanges making the cut.

The table below lists down the percentage of header bidding sites each vendor was in by year:

20172020Change
AppNexus (Xandr)62%72%16%
Index Exchange56%62%11%
OpenX51%50%-2%
Amazon51%72%41%
AOL47%30%-36%
Rubicon Project42%59%40%
Criteo36%33%-8%
Sovrn35%26%-26%
Sonobi25%13%-48%
DistrictM13%18%38%
TripleLift9%20%122%
2017 most found bidders
2020 most found bidders
change in bidders

As a whole, apart from Sonobi dropping out and TripleLift entering, the top 10 ad exchanges remained intact. But there was a substantial movement within the top 10.

Amazon saw the most improvement – jumping from 4th with 51% to 1st with 72%. Otherwise, it was a mixed bag: half of the vendors saw increases, half saw decreases.

What does this data mean?

Our expectation was that all exchanges would have seen a lift. After all, with the number of bidders per header bidding site increasing from 4.9 to 7.2, it would make sense that every vendor would see a bump as publishers sought more demand integrations.

That said, there are still latency risks with loading up a wrapper with bidders (as well as engineering time to integrate), and vendors have to make a case why their demand is worth it. It’s possible that some of these older players – such as AOL (which saw a 36% drop) are having trouble making that case, leading to drops in usage.

One caveat: since we aren’t able to track server-side calls through Amazon’s Transparent Ad Marketplace and other S2S solutions, it’s possible that this mix changes when you include backend calls.

Summary 

Looking at three years of data, one could make the argument that not much has really changed: header bidding adoption is up only by 14%, the top 10 adapters saw only one bidder drop-out, and no client-side wrapper made any dent in Index Exchange’s monopoly.

But we also see more to the story. With a doubling of unique adapters found and a 50% increase in the number of bidders per publisher, it’s clear that the past few years has been less about new publisher adoption and more about the ad networks and ad servers evolving their products to be header-bidder-friendly.

Amazon’s rise to #1, moreover, cannot be overlooked and cements their status as an ad tech giant.

As we look ahead, what does the future hold for header bidding? We have a few guesses:

  • Header bidding adoption will max at 83%-85%. Even Ads.txt – which is a requirement for some ad exchanges – has only a 92% adoption in the US. Some publishers seem reluctant to make the leap.
  • The number of bidders per client-side wrapper will start decreasing at some point, for a few reasons:
    1. Latency concerns will lead to hesitation about new bidders
    2. The incremental revenue impact of new bidders will be minimal, leading to apathy toward new integrations
    3. These client-side endpoints will move to the server.
  • The top 5 bidders – Amazon, Index, Rubicon, Xandr (AppNexus), and OpenX – will continue to thrive, and it will be tricky for other vendors to upsets these rankings.
  • The client-side wrapper market will continue to be fragmented, with Index Exchange maintaining its #1 ranking and no one else getting above 5% market share.

This is a guest post by Chris Shuptrine. He has worked in ad tech for over twelve years in a variety of roles – customer support, product management, and marketing – giving him perspectives from both the advertiser and publisher sides. He’s currently the VP of Marketing at Adzerk.


Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.