Combinatorial Bidding: What It Is and How It Delivers 40% Lift in CTR for Our Clients


Imagine you’re at the grocery store, and before you drop a can of beans in your cart, you analyze tens of thousands of factors to determine if it’s the ideal can of beans for your recipe — and if it is, what the appropriate amount to spend on it would be. This isn’t the stuff of fantasy — not in programmatic advertising anyway; it’s a process known as combinatorial bidding, which uses to identify the optimal impression and identify the appropriate bid price according to a particular client’s campaign goals.

Location, time of day, device type, messaging — these are just a few of the literal tons of data points we weigh using combinatorial bidding, allowing our advanced programmatic platform to determine, in real time, the most likely candidates to click or take action on an impression. This process also allows us to weigh key campaign performance insights to continuously update our models throughout the life of the campaign, optimizing for high marks on delivery metrics like CTR and CPA.

Focusing on the impressions with the highest potential for performance.

For an audience to take action on an ad, they need to see compelling creative at the right time in a place where prospective buyers are most receptive. This may seem like common practice in the ad world, but — and here’s where combinatorial bidding comes into play — is able to analyze far more factors to determine the best times and places to deliver these impressions, thanks to the unique tandem of combinatorial bidding with unstructured data.

Our models are always learning and adapting in real time, based on their current effectiveness within that specific campaign — or similar campaigns, in order to optimize models more quickly. And with unstructured data points, we have the flexibility to determine which factors (time of day, location, etc.) will have the most influence over campaign performance, thus optimizing impressions from the targeting stage through delivery. In essence, a better-informed targeting strategy begets a more receptive audience, which means campaigns are better tailored to meet the unique goals of each client.

Optimizing ad spend by valuing bid prices.

We see about 300 billion, with a B, bid requests each day here at, which gives us a lot of data to analyze. And for every single bid, our models use this data to score the desirability of each impression on a scale of 1 to 10. Known as deciles, these numerical values allow us to determine the appropriate bid price for that particular impression, with higher deciles commanding a higher bid price, and a lower bid or no bid at all for lower deciles.

For instance, if our models show, based on past impressions for a user (or similar users in other campaigns), that this individual is more likely to click or take action in a certain location or at a certain time of day, that impression would have a higher decile than other locations and times of day, meaning we’d be willing to bid more on that impression if the circumstances called for it. But a lower decile doesn’t necessarily mean low value; just like the value of a low-cost house can exceed its market price, an impression can hold more worth than a decile might indicate. Bidding on these low-cost/high-value impressions allows us to secure those more desirable clicks — a combinatorial process that ensures quality performance and lower CPMs for our clients.

Fulfilling campaign goals and driving performance.

There’s a reason combinatorial bidding is a fast-growing trend in the programmatic space, and why it’s been our go-to bidding methodology for years: It delivers, as the 90,000 daily active campaigns in our platform can attest. When measured against a random holdout set that doesn’t use the model, our bidding methodology delivers upwards of a 40% lift in CTR for our clients’ campaigns. And it’s because we leverage unstructured data, as opposed to audience segments, that we’re able to achieve this level of granularity, flexibility, and performance for our clients.

Everything from the billions of bid requests we receive to each element-level factor is deposited in our platform as unstructured data, empowering our models to sculpt this wealth of information into the most efficient and advantageous combination. It also allows us to optimize the bid price at the impression level, while other programmatic platforms can only optimize at the audience level.’s ability to target, bid, optimize, and report with the breadth and detail of unstructured data is unique among programmatic advertising solutions, and provides our clients with an unprecedented level of insight into the effectiveness of their campaigns.

For more about how our combinatorial bidding process can help you achieve your marketing goals, get in touch at