Last week, in “The Ingredients of Our Secret Sauce: Part 1
,” we began to break down some of the key technology ingredients that power the Turn Platform. Although we can’t publicly talk about the recipe to our secret sauce, it’s helpful to understand what those ingredients are. We talked about the importance of data and machine learning-based optimization, but what other ingredients are needed to drive performance?
Today, Turn considers billions of bid requests every day. We are architected (you’ll see a theme here) to horizontally scale to any number of billions of bid requests, and do so by using each of our automated algorithmic resources as efficiently as possible. This is important because scale matters. It’s one thing to look up all of the data for an impression and calculate the optimal bid response, but it’s another to do it at scale for thousands of campaigns and billions of impressions per day. If your servers slow down, even the slightest bit, they won’t respond to the RTB auction in time, causing advertisers to miss opportunities to bid on relevant impressions (i.e. loses campaign reach).
When it comes to server latency in an RTB-based environment, there is a set time period, generally 75 to 150 milliseconds, to respond to a bid request from each ad exchange. This is the round trip time required from the exchange’s servers to Turn’s servers and back again. As this round trip involves communicating over the Internet, once the bid request reaches Turn’s servers, we have a lot less than the 75 – 150 milliseconds to respond to that call. Yet, despite all of the real-time data retrieval and bid optimization algorithms we utilize for each impression, Turn’s bid processing is in the milliseconds, which is an achievement that speaks to efficiency.
The real proof in the pudding is when we receive results from our exchange partners and media buying clients. Our partners show error and win rates; hence, Turn’s scorecard tells us what % of their inventory we bid on. According to Google, in Q4 Turn’s successful bid response rate was over 95% for the AdX exchange compared to an industry average of 59%. That’s a huge difference, which translates directly into campaign reach and performance. Every time a DSP can’t respond in time to an RTB auction results in a lost impression opportunity for the advertiser.
The second proof comes from our clients. We are not about cheap ‘cookie bombing’ inventory tricks; we are about meaningful results in areas like data warehousing, analytics, and audience insights.
What’s important to take away from this is that it’s not about one technology ingredient; success comes from blending them all together. When you combine a highly scalable infrastructure with data management, algorithmic optimization, and machine learning models you have a technology platform unlike any other. Data-driven advertising is at the core of Turn’s technology and we are committed to building innovative solutions for our agency and brand customers.
If you have any immediate questions, feel free to contact me at Dominic.firstname.lastname@example.org
VP of Engineering