Header Bidding: Limitations and Considerations
Has there been a bigger topic of discussion over the last year than the rise of header bidding? Since it comes with a lot of promise, much of the conversation has been around the benefits it brings publishers, but those of us who are knee-deep in header bidding are just as concerned with its limitations, and developing solutions that will enable us to get even more out of the technology.
Here are three areas we’ve spent significant time discussing at Freestar as it relates to header bidding complexities and achieving success for our clients. (And if you’re looking for a more general post about header bidding, see my previous post here).
If you’re intentionally stopping a page from loading so you can go out and get bids, you run a risk of increasing page load time, which means that in your effort to make more money, you’re creating inefficiency.
As a publisher, the key is to balance the benefits and the drawbacks. Are you making more money than you were before, at the cost of losing some of your slower connection users? And those visitors who weren’t able to load your site at all…will they come back? We like to look at both of these in our consideration of where that balance is.
One solution we’re implementing to reduce risk of latency is asynchronous loading of our header script. Unlike synchronous loading, which stops everything on the page from loading while bids are called and returned, this method lets everything else on the page load while you’re making requests, but holds the ad server from loading. It’s been a big improvement for us, but is not yet commonplace since it requires a specific level of knowledge to implement.
CONCURRENT CONNECTION BOTTLENECKS
Mobile phones are already overtaxed for bandwidth with what they’re trying to download, and header bidding is asking them to do a lot more.
Based on your browser, you only have a certain number of concurrent connections per domain that you can make. For example, an older Internet Explorer browser may only support six concurrent connections. If you’re requesting 18 header bidding partners, who gets those six connections? A newer Chrome browser may have 14 concurrent connections, which is better, but even then you’re still maxing out.
This leads to a number of questions on how to solve an issue that header bidding has created. Talented data scientists are needed to make advancements in this area. Which are the best partner connections based on a number of variables? How many connections are ideal? What timing is optimal? Header bidding has dramatically increased the need for skilled analysts who can look at complex data and uncover the most finite factors that will lead to success. Qualified talent in this area is scarce, and hiring is competitive, which means publishers should make it a priority before the need is critical. For publishers that aren’t big enough to have an inhouse data science team, this will start separating the haves from the have-nots. If a data science team isn’t in the budget, serious consideration should be given to working with someone that does this kind of optimization.
CONFLICTS OF INTEREST
Header bidding doesn’t necessarily come with fair collaboration or transparency. This is nothing new in ad tech, of course, but it’s a more crucial issue than ever as data collection becomes more sophisticated and providers become more intertwined.
As publishers enter into relationships where they’re providing valuable data to service providers, these providers often reap the benefits of the data by using it to get visibility into the market. That leads to a clear conflict of interest and many publishers left with no leverage if the partner is also providing buy-side services or doing its own buying.
A stronger commitment to education must be required, with publishers consistently asking themselves important questions about who they’re doing business with. (What are this provider’s policies around proprietary data? Do they have my best interests in mind? Will they deliver an expected level of transparency?) If you can’t get answers, it might mean the risks outweigh the benefits.