The inventory flowing through your header tag needs to undergo constant optimization to ensure that you’re maximizing revenue and brand participation in your market.

 We achieve this optimization by using a genetic algorithm “GA” to study bid data. The GA is a solving model that mimics the Darwinian principles of natural selection, where only the fittest members of a population survive the competition for limited resources. Pricing ability in header tag implementation is largely driven by the price levels defined in an ad server’s line items. Each of these price levels are treated as candidate price strategies, and the GA optimization is essentially submitting these candidates to competition.

 Unlike traditional waterfalls where partners are allocated inventory to monetize, header tag only accepts the impression if it offers the best price within your ad stack. This competitive environment adds an interesting dynamic to optimization problems, and requires a more sophisticated approach to pricing inventory. Essentially, the GA learns how to maximize the value of your ad impressions so that you don’t have to.

 Making sure the GA is setup for success

When properly configured, dynamic allocation “DA” (formerly Enhanced Dynamic Allocation or “EDA”) will provide strong incremental demand critical for publisher success. With that said, header tag will only work as well as its ability to participate in a properly configured auction. Modern ad servers equipped with DA provide critical reporting information that must be studied. The most important indicator is the clear price versus the custom criteria targeting parameter. Whenever there is a delta to the negative, it means an impression is clearing at a sub-optimal CPM against DA and may require configuration changes.

 An example of this would be where a header tag submitted a bid of $8.00, but it was then monetized by DA at $4.50. This signals back to the GA that $8.00 is not a competitive rate, when in reality it is a function of ad server decisioning instead of true price competition.

From data collection to constant refinement of price selection and inventory exposure, the GA optimization process never ends. In an ever-changing and evolving bid landscape, even the highest performing portfolio needs to learn to adapt. Using empirically-driven and revenue-oriented evaluation methods, the strongest strategies are continually rewarded with more inventory allocation, while the weaker are reigned in with decreased inventory exposure, or even elimination from the strategy. Using these techniques and others we ensure optimal performance with the adaptability required for the long term.


  1. In the “clear price versus the custom criteria targeting parameter” example, why would the clear price be less than the targeted parameter dictates? Would this be only in situations where the you have less granular line item price buckets (versus .01 increments) or if the values set for the targeted parameters are purposely inflated in the stack?

    Thanks for the great articles!


    1. Great question and thanks for the positive feedback! This is related to price level granularity where we rarely go down to 0.01 increments so there will be some level of rounding down going on.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s