ByBen Kubota, writer at
I'm saving the world at night, that's why I'm always so tired in the morning. product guy at, talk to me via twitter @omdb.
Ben Kubota

The way you sell your inventory is in the midst of an evolution - once again. With Header Bidding maturing, this approach might simplify and hopefully improve the way advertisement is displayed on your site.

But before we go into the advantages of Header Bidding, lets acknowledge, that the current way on how classic display advertisement is affecting the user experience - especially on mobile - is dreadful. No matter how much energy we spent in optimizing the site speed, on thing is for sure, before the last ad unit is displayed, more then 10s have passed. And this is just because we're all still relying on a legacy technology.

A brief history of display advertisement

In the early days of display advertising, publishers worked with advertising networks to monetize their unsold ad placements, i.e. remnant inventory. The advertising networks, in turn, acted as middlemen between Fortune 500 companies who wanted to spend their advertising budgets online and publishers seeking ad revenue for their editorial content.

In this process, the publisher first negotiated a contract with the advertising network for a specific campaign. The contract usually entailed a fixed CPM price in exchange for a specific amount of ad impressions. When the paperwork was completed, the publisher received an iFrame or JavaScript tag from the advertising network to display the creative through an ad server. Since the CPM price was agreed upon beforehand, publishers didn’t spend much time optimizing the delivery of the campaigns because prices were typically fixed.

The relationship between the publisher and the advertising network wasn’t a technologically savvy interaction. Instead, the biggest issue with this approach was business related. Hence we see a lot of data being transferred back and forth for every single ad that is being displayed.

Why is that? It was problematic for the publisher to forecast the amount of remnant inventory (as opposed to direct sales) that would be made available leading into the month. Forecasting was further exacerbated by natural fluctuations in website traffic from month to month. Publishers typically faced one of two scenarios: under-delivering on impression goals and agitating the ad network, or under-selling impressions and losing monetization potential. Both scenarios involved many emails back and forth adjusting impression goals. A highly inefficient process, to say the least.

As the industry started to mature, publishers realized they weren’t necessarily making the best deals or finding the right buyers for their remnant inventory. So in 2010 a shift towards programmatic advertising started to take place. By automating parts of the buying and selling experience, programmatic eased price negotiations and created more value for the publisher.

A brief overview on programmatic advertisement

Programmatic ads enabled publishers to move away from advertising networks and manual insertion orders in favor of Real Time Bidding. Computers essentially replaced the human interaction involved in buying and selling digital media. This meant that within the blink of an eye, a publisher could now sell an impression to any buyer using a defined set of business rules, such as pricing floors, which both parties agreed upon. While this model greatly reduced time spent on planning and forecasting, it didn’t give publishers all the tools necessary to maximize yield.

While Real Time Bidding makes it possible to efficiently auction off an impression to the highest bidder or buyer in one marketplace, the underlying technical infrastructure remained unchanged. Previously, the technical function of the iFrame or JavaScript tags was to deliver creative campaigns with a fixed CPM price from the advertising networks. With the development of the new bidding process, their job is now to deliver creatives from winning bids in the marketplace. So the task changed, why hasn't the technology?

To be fair (and to simplify), there are two sides to the story. There is the supply side of things (the publisher, offering inventory - or SSP) and the demand side (the advertiser, the agency, the buyer - or DSP). We've seen a lot of excellent developments on the DSP side, but the SSP doesn't seem to follow. Primarily because the waterfall is still in effect.

The issue that arises is that the price from the winning bid isn’t known to the DSP (the ad server)*. This makes it difficult to optimize yield and effectively curtails the publisher by having to manually adjust and optimize the cumulative CPM price in the ad server based on historical data. Consequently, the allocation of impressions by the ad server is an indicator of past performance rather than future possibilities.

In a typical scenario, the publisher works with multiple programmatic vendors to maximize revenue performance as not all marketplaces perform equally well. The obstacle here is when the best performing marketplace has run out of applicable buyers, it can’t effectively communicate this to the ad server. Publishers, therefore, try to anticipate performance by setting up the iFrame and JavaScript tags from the programmatic marketplaces in a “trickle-down” or “daisy-chain” system resembling a waterfall.

In each position of the waterfall, the publisher manages how much traffic should be allocated to each tag. If the publisher takes a conservative approach, valuable impressions might be left on the table because not enough traffic was sent to achieve premium rates. If too much traffic is being sent, on the contrary, we encounter the opposite issue where ad requests aren’t turning into paid impressions. To minimize the damage of the latter scenario, programmatic vendors allow publishers to create “passback” tags in the event that the marketplace can’t turn the ad request into a paid impression.

In a sense, passbacks are the workaround for a workaround of a missing feature. Passback tags send the ad request back to the publisher, which then can be re-directed to the publisher’s second best option. The difficulty is that the passback selection isn’t a lucrative option for many advertisers. It’s basically the unwanted impression that all buyers passed on in the first segment of the waterfall. The passback option is far from an optimal solution and has precipitated another stalemate in the industry. And most importantly, it takes a lot of time to figure out, who should get the impression, which impacts the user experience.

To summarize, the transition to programmatic ads using Real Time Bidding helped automate the business side of buying and selling media, but it only marginally improved the technical aspects of it because the integration with the ad servers remained unaltered.

A new era in display advertising: Header Bidding.

In a nutshell, Header Bidding automates the technical aspect of buying and selling digital media by eliminating the need to manage 1) price optimization, 2) the waterfall, and 3) passback tags. With Header Bidding, we’re also able to isolate every single impression on the publisher’s website in an effort to maximize yield.

How does it do all that?

With traditional display advertising, the ad request originates from the iFrame or JavaScript tags in the ad server’s waterfall. Inefficiency is encountered when a second tag in the waterfall might be poised for a huge campaign that matches your site’s verticals and performance metrics, but if the first-option tag hasn’t reached its goals yet, the inventory will never be exposed to the buyer.

Header Bidding, on the other hand, handles this situation much more gracefully as it moves the ad requests away from the ad server and onto the publisher’s website instead. As a result, when a user visits the page hosting the editorial content, the Header Bidding technology is now able to simultaneously call all programmatic buyers at the same time when auctioning off the impression. Since more buyers are able to partake in the auction this way, publishers also generate more revenue due to the increase in competition.

Another aspect of Header Bidding that simplifies the workflow for publishers is that the need for passback tags becomes obsolete.

Header Bidding is, in that sense, a tagless approach to buying and selling digital media. If a buyer doesn’t want to bid on the inventory, the advertiser doesn’t need to send back the impression to the ad server any longer. From now on -- in real time -- they simply don’t partake in the auction. This leaves the door wide open for a different buyer to entertain his or her luck.

What are some of the caveats to Header Bidding?

One has to acknowledge that the technical barrier to entry is more difficult compared to the tag-based solution. With the latter, we’re simply copying and pasting a piece of code to the ad server whereas a developer is required to implement Header Bidding technology. Not all publishers have the resources readily available to hire a developer for this purpose.

Another concern is that with Header Bidding, rather than calling each advertiser in successive order, we’re now calling multiple marketplaces at the same time. This might cause the page to take longer to load or it might even load faster, depending on the number of partners you work with. It definitely increases the number of open connections, which is an issue on mobile devices. That said, each publisher needs to figure out how much latency they’re willing to accept when chasing the perfect media sale.

We’re currently figuring out what is the right number for us. Early results show an increase in CPM (up to 30%) and no negative effect on the page load time. Hopefully this is a step in the right direction for the advertisement industry to catch up with current publishing trends. And this can only be the first step, to fix the user experience. Maybe, if we had this a couple of years ago, slow loading websites wouldn’t have paved the way for Google AMP and FB Instant Articles.

*) This isn't true in general, as some DSPs include their own RTB SSP, like DFP integrated AdX with their "price priority" setting, but it still doesn't allow you to plugin several SSPs and give each of them a fair chance. At least not without a lot of effort.


Latest from our Creators