All Categories > Header Bidding

Under the Hood: Study Break Media's Header Bidding Implementation

Study Break Media (recently acquired by Chegg) reaches 85% of students in the United States through their sites EasyBib, CitationMachine, BibMe and CiteThisForMe. With over 99 million unique users and 1.34 billion users per year, they have invested heavily in programmatic advertising to monetize their inventory, 95% of which is considered remnant.

Inside their header auctions

The Study Break Media team was an early adopter of header bidding, setting their first parters live in Q4 2014, reaching 4 partners by mid 2015. They now have 10 total header bidding partners, including Sovrn, running through their customized Prebid.js wrapper. Sovrn and three other top partners are integrated outside of their wrapper through an API integration.

“We chose to use Prebid for our wrapper because it allows us to integrate and test new partners rapidly.”

Before header bidding, Study Break Media’s DFP waterfall purchased around 60% of their inventory. After header bidding, 50% of inventory is won at a higher CPM than waterfall. They consider the 50% revenue attributed to header bidding as adding value to their overall solution, All demand is set up at Price Priority in DFP, so header bidders will only serve if they are able to monetize at a higher rate than their existing setup.

Currently, most of Study Break Media’s inventory is sold via header auctions save for a few remaining waterfall partners. They are still gathering an analyzing data to determine if header bidding valued demand is there to warrant an eventual move of all inventory to header auctions. The main roadblock from a full transition to header bidding is the lack of a solid solution for dynamic flooring based on page session depth. With high engagement and user sessions that reach up to nine pages and deeper, there is a noticeable drop off in bid density from partners on the deepest sessions. Remaining waterfall partners are in place to pick up these impressions.

Customizations

Each of SBM’s header bidding partners are only able to asses performance based on their own numbers. With permission, they with share SOV, discrepancy reports and KVP line item reports from partner to partner. This allows their team to put context to partner performance and explain the true value of each opportunity. Each customization that is implemented is vigorously A/B tested. Recent test have included partner density and partner timeouts.

More data, more density, more testing

While header bidding has not been a perfect solution, it has alleviated quite a few pain points for the SBM team. The most obvious of which is a drastic improvement in bid density and auction pressure, producing accurate valuations for each impression. But the real nuance and intrigue comes from the mountains of data that are now at their Ad Operations team’s disposal.

  • Access to log-level bid data in real-time has been huge for making informed decisions, faster. The days of blended and cumulative data with a 24 hour lag are gone.
  • Partner-level discrepancy and performance metrics make partner evaluation straightforward.
  • Instead of having to duplicate entire ad stacks, tags and defaults to conduct A/B testings, their team can run true A/B testing environments at scale with KVP reporting in DFP.
  • Header bidding effectively removes the need for managing and accounting for multiple impression passbacks.
More money

Study Break Media has seen higher CPMs from header bidding partners across the board. Additionally, adding partners has increased the volume of their inventory that is purchased through the header. Interestingly, header bidding has also changed their relationship with partners. Instead of focusing on waterfall placement and flooring, they are empowered with data to dig into specific discrepancy and bid behavior issues.

Never miss a post.

Subscribe and get updates on new blog articles,
weekly curated industry news, and more!
  • This field is for validation purposes and should be left unchanged.