Understanding Cookieless Measurement

Running a successful marketing campaign requires analyzing the outcomes of your campaign and how the decisions you make during the campaign affect those outcomes. Are ads on a big publisher or a collection of small publishers producing more new clients? Is creative A or creative B resulting in more interest in your product? By measuring these and many other results, you can make changes to your current and future campaigns to win the most business. 

Quantcast leverages many kinds of data, including the traditional third-party cookie and other more probabilistic pieces of data.

Attribution

Attribution is a key measurement used in advertising. Which impressions (if any) led to conversions? With third-party cookies, there was a direct link between the impression and the conversion. This isn’t necessarily true with cookieless measurement. Without the third-party cookie, you have just a set of separate impressions or conversions with no link between the two. Thus, you do not have the measurement needed to re-allocate your budget.

Deterministic and Modeled Data

So, how do we measure attribution without third-party cookies? Two common ways today are via probabilistic modeled attribution or deterministic identifiers (such as email addresses). These methods require collecting data about the impression and the conversion event. For the impression, much of the data comes from the publisher, such as if the user was logged in, what their logged-in email was, where and when the ad was served, and many other characteristics of the ad. For the conversion, the data comes from the marketer, typically from a pixel placed on the marketer’s website or via a conversion API that the marketer’s system uses.

Deterministic conversions are highly accurate; however, they require logged-in users on both the marketer and publisher sides, and they need to share the email or a common ID based on the email (such as Liveramp ID or UID 2.0). As a result, there is much less inventory available that can be used for deterministic attribution, and that inventory is more expensive.

Modeled conversions are less precise, but they can be measured on much more inventory. Since these conversions are modeled, we can aggregate the scores to become more accurate as campaigns increase in size. While measuring a single conversion will depend a lot on its specifics, we can say with more confidence that a campaign drove 50 or 500 conversions when we aggregate our scores together.

At Quantcast, we combine the two: We use deterministic approaches when we have the necessary deterministic data and modeled approaches when deterministic data is not available. This allows us to get the best accuracy and, at the same time, not restrict what inventory we can purchase. Clients report that our measurements match theirs (when they are using other cookieless measurement solutions). We also validate by comparing to situations where we still have third-party cookies or other identifiers.