Header bidding is a relatively new concept, yet it is having far-reaching and dramatic consequences on the industry (see Header Bidding, Google Server-Side Header Bidding and 2016-12-11 Amazon's Header Bidding for related articles). As any major theme becomes more of a mainstay in our industry, the IAB tends to step in and, as the predominant standard-setting body in our industry, tries to set some standards. Header bidding followed the same course and the IAB just released a draft of its attempt to standardize header bidding wrappers, which is available here: https://iabtechlab.com/specifications-guidelines/header-tag/ (on page 2, you can see that TripleLift participated in this exercise). Very few actual rules are provided, the draft is mostly a set of suggestions about what ought to considered and accomplished. The draft covers a few focus areas, which are line item setup, data, and reporting & discrepancies. Each are discussed briefly below.
The premise behind header bidding is that a “dormant” line item is created in the adserver. For most adservers, a key-value specified in the header, which then triggers the appropriate line item in the adserver (those that support header bidding natively don’t necessarily require this). The recommendations from the IAB include:
- Some ad servers have limits on the number of line items, so do not create an unreasonable number of line items. Also consider appropriately-descriptive names.
- It’s possible to either have all buyers share the same line item(s) or use their own
- This means using less price granularity if bidders (buyers) are set up separately
- You should support a number of distinct buying methods, including:
- Direct guaranteed: this should compete at a very high level in DFP (e.g non-sponsorship direct reserved guarantees)
- Indirect non-guaranteed: generally pure price competition, below specifically prioritized campaigns
- Direct non-guaranteed (i.e. PMP): prioritization in the adserver should align with the what was sold, which can be highly variable
Data and reporting
- Rules are defined for certain key-value pairs in the header bidding setup. This includes: bidder (this is the buyer); winning ad ID; winning bid amount; and placement size.
- The relevant data to report on is: winner within each bidder (header buyer), who won the container’s mediation, and who won the logic within the adserver. Some others require additional data, like video and in-app, because of caching etc.
- Discrepancies are an important consideration. Each new buyer should have discrepancy testing for revenue, impressions, and key-value, and should be repeated with some frequency.
- Partners should work to have an agreed-upon discrepancy resolution mechanism.
Nothing in this draft is particularly controversial. The approach that’s being taken appears to be, in the first instance, asserting a fairly high-level consensus among the header players and creating at least a base-line of understanding. It’s likely that in future versions, suggestions about considerations may evolve to be requirements - and standards may be set around how certain things may be done. But this was the first step of taming what has largely been a completely ad-hoc and “wild-west” set of implementations. It is likely that the technology behind pure header bidder wrappers will quickly become commoditized and value will move to other parts of the food chain.