We wanted something simpler, easier and more powerful.
Despite Prebid.org's claims that their wrapper requires no coding and that it's simple to setup, we've had many problems using it:
- We saw revenue drops of over 30% when migrating from one version to the next.
- We tracked down bugs in the selection of winning bids.
- We struggled with the initial implementation to get header bidding to work properly with Google's ad server and get the correct bids and targeting to get through.
- We struggled finding the right timeouts only to figure out via testing and measuring that timeouts are largely irrelevant and device and network dependent.
- We saw race conditions with header bidding auctions not running at the right time and not sending winning bids to our Ad Server
These problems forced us to dig into the Prebid.org based wrapper and a create a set of tools that would make our ad operations easier, faster and more profitable.
- We thoroughly tested our Prebid.org implementation and quickly determined that the wrapper wasn't doing the right thing.
- We forked our own version of the Prebid.org code and added fixes to chose the correct winning bids and send the correct Ad Server parameters.
- We built a reporting system that would let us compare GAM numbers against our Prebid.org partners' impression and revenue numbers
to make sure we were getting paid for all the impressions we served.
With our forked wrapped we were able to try many different revenue maximizing ideas that added up into a significant revenu lift.
- We increased revenue by using a more profitable backfill strategy.
- We increased revenue by setting long timeouts and our own timeout handler.
- We increased revenue by making sure GAM Ad Server auctions were getting header bidding auction data every time.
- We increased revenue by making sure we were only bidding on viewable ad placements.
We started using our data to run experiments and to optimize our ad operations
- We were able to reduce unfilled rates and maximize our yield by using some innovative pre & post render, backfill, refresh and 2nd auction boosting strategies.
- We added a layer of protection to prevent malicious ads from redirecting our users and reducing our overall impressions per session.
- We added tools that made it trivial to add new partners and run tests.
- We set up custom bucket granularities in GAM to maxmize bid prices in Ad Server auctions.
We made these small incremental improvements and measured their impact. We kept each feature that yielded a small boost usually 5% - 10% over the previous iteration.
We settled on a set of features that gave us a 30% increase in yield on average. On some pages 5% on others 50%.
Now we're offering what we learned in an easy to use tool. Give us a try. You can even compare us against your current Prebid.org implementation by using our analytics and pbjs debugging tools.