Scaling Ad Operations
Sortable Ad Engine helps publishers automate their ad operations allowing them to focus on creating great content, and stop worry about ads. As the team took on more sites and more impressions, it became overwhelming to limit lost impressions and discrepancies through updating line items, price floors and passbacks. As Daniel Jantzi, Head of Ad Ops recalled, “As we were managing more sites it got more painful and we were literally throwing money out the door every day.” The Sortable team was dedicating significant manual maintenance to minimize the 5-15% discrepancies or impressions lost for each ad request and up to 20% discrepancies for each campaign.
Sortable looked for a way to stop losing revenue from lost impressions in a traditional waterfall and ultimately increase yield. They started by testing header bidding integrations within their Ad Engine.
With multiple header bidders in their Ad Engine, they didn’t have to manage price floors and could scale out across ad networks without worrying about line item updates. Less manual maintenance and decreased loss was a double win.
Removing the Guesswork
Sortable integrated Sovrn’s Header Bidding API within their Ad Engine, along with several other ad networks. Compared to other header bidding solutions, “Sovrn dug into issues that got things fixed – we reduced the discrepancies to record lows. We have very thorough weekly meetings with our account manager – to work with us to get the numbers up and alleviate issues. You guys made it work,” mentioned Daniel.
Below are some questions we asked Daniel about their header bidding integration:
You mentioned that your previous discrepancies were between 10-20%. By how much did header bidding reduce lost impressions?
Header bidding doesn’t really reduce discrepancies, but it does reduce loss because there aren’t any passbacks involved as there are with tagged solutions. Each header bidder partner has discrepancies in reporting, and it’s usually similar to the tagged solutions, just minus the loss in passbacks. We were pleased that Sovrn’s bidder solution had significantly less discrepancy than other partners. There are still header-bidding partners that have between 10-20% discrepancy in our stack, but for us, Sovrn has one of the lowest at around 4-7%.
How long did it take you to implement header bidding with Sovrn compared to other vendors?
The actual programming to conform to a partner’s API takes anywhere from an afternoon to several days. In our experience, most of the time in setting up a new partner’s header bidder is in the communication, e.g. sending API docs, legal agreements, generating tags, checking test pages, etc. The entire process can take a month, or get stalled and last even longer. With Sovrn we went from initial call to running on 100% of approved inventory in less than two weeks. The responsiveness of the people at Sovrn (Chris Crawfurd, James Manor and Tony Casson) made it so we were rarely waiting on a response.
Were there any specific aspects about our API that you liked? You mentioned efficiency, but can you unpack this for us?
Sovrn’s API is straightforward – JSON bids go in, JSON prices and creatives come out. It’s an implementation of the OpenRTB standard, and their examples are easy to follow. They’re not unique in that regard, but you really don’t want to be unique when implementing a standard. Regarding efficiency, some vendors require a separate http request for every ad unit on a page. Sovrn’s implementation of the OpenRTB protocol allows all the ad units to be batched into a single request (as it should be), avoiding some overhead. It’s not a huge difference, but every bit counts.
Sovrn’s average server response time is also among the fastest we observe, so we can’t ask for much more in that regard.
