Match Rates Are Just A Number

terrychenData-Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.

Today’s column is written by Terry Chen, product manager, Adobe Analytics Cloud.

I recently overheard a few co-workers debating the merits of online dating apps.

One was particularly enthusiastic about the demographic and geographic filters that can whittle down the field of potential matches and fondly recounted a past courtship of someone more than a few years senior. The other co-workers were not impressed, uninterested in dating someone outside of their age group.

“Oh, please,” the first responded. “Age is just a number.”

Advertisers, too, are obsessed with finding potential matches and using data to fine-tune their approach. Most major brand advertisers already use or are in the process of signing up with a data management platform (DMP) to get more value from their customer data. With first-party data, bigger is better – the more data an advertiser has, the bigger their particular audience segment and the better the scale they can achieve when they try to reach that audience across devices, apps and websites.

This pursuit of data targeting at scale has popularized the concept of match rate, which usually refers to the percentage of users within a given audience segment that a DSP recognizes. While this is an easily calculated metric, it is a poor proxy for the question marketers actually care about: What percentage of an audience segment can advertisers actually reach with their ads?

Data naturally atrophies – or leaks – throughout the various steps within the activation progression. Lost profiles in the activation process directly correlate to decreased match rates, reducing the targeted scale advertisers can achieve. The graphic below shows how profile loss occurs from the first step of onboarding offline data all the way to the last step of actually reaching the consumer.



As the graphic illustrates, marketers must be extremely specific when they refer to match rates because there are four separate matches that actually occur throughout the process.

  1. Offline -> Online onboarder
  2. Onboarder-> DMP
  3. DMP-> Demand-side platform (DSP)
  4. DSP-> Supply-side platform (SSP)/inventory

The effectiveness of the first two matches relies on the pre-existing overlap in footprints between the offline data, the data onboarder and DMP; marketers cannot simply demand high match rates.

However, the third match – syncing the DMP’s audience to the DSP’s audience – is what most marketers refer to when they talk about match rates. And while leveraging a DMP that is closely integrated with a DSP is the most immediate thing marketers can do to boost their match rates, it’s the fourth step – matching audiences at the inventory level – where the rubber meets the road.

Instead of starting with a minimum match rate benchmark, marketers should start by determining which channel and device their audiences are on and evaluating inventory availability for each ad format – or, increasingly, which set of screens and formats – they intend to deploy.

For example, a large media network with ad buying capabilities may see the highest match rate overall but it is most likely finding most of those users across its vast display inventory. If an advertiser is concerned about branding and seeking premium, large player-size video inventory, then it would need to find a partner that would be most successful in reaching that goal.

When using a traditional performance DSP, marketers will have few issues reaching users across display inventory, but their large player-size video scale will be limited. In this situation, marketers may have a lower overall match rate, but they will have more scale across large player-size video inventory, as well as other channels like social.

Looking at match rates in the strategic context of the overall ad buy, rather than as a KPI, is the most important thing marketers can do. They should also insist that partners share the formula they use to calculate match rates as each vendor can calculate them differently due to lack of industry standards. Certain mobile devices or browser types, for example, follow strict protocols regarding tracking using third-party cookies. Beyond just employing a DMP that is tightly integrated with a DSP, marketers must think about match rates at the inventory level, considering how device and channel affect ability to deliver.

And while delivering a target audience at scale is important to maximize data’s efficacy, data portability is even more important. The best match rates in the world don’t do any good in a closed ecosystem that prohibits data from being applied elsewhere.

Looking forward, it’s not hard to imagine a universal identifier as the pickax that finally breaks down walled gardens. It’s also not hard to imagine artificial intelligence-driven audience discovery and segment creation leveraging that universal ID.

In the short term, however, if you hear someone talking seriously about match rates, remind them: It’s just a number.

Follow Adobe (@Adobe) and AdExchanger (@adexchanger) on Twitter.

Editor’s note: An earlier version of this column erroneously attributed the data in the graphic to Nielsen Catalina Solutions. The figures reflect internal Adobe estimates.

Enjoying this content?

Sign up to be an AdExchanger Member today and get unlimited access to articles like this, plus proprietary data and research, conference discounts, on-demand access to event content, and more!

Join Today!


  1. I have to disagree. The best approach isn’t to start with segments and devices or channels, but the individual. If you can identify a person across channels and devices, you can choose the channel that they prefer and/or are most likely to convert. Problem is, you can count the number of companies that can do this at scale on one finger.

  2. You’re right, Terry … match rates are just a number, but for the reasons you point out, a poor performing number that not enough people really understand. In my opinion, we as an industry too often focus on the numerator (the 20M UUs you refer to above) not the denominator (the 100M UUs you started with). An industry-standardized ID would help with this tremendously, perhaps even eliminating the term “match rate” (which assumes much data loss) and allowing the industry to focus on audience overlap (which assumes no data loss).

  3. It is definitely the case that marketers seek the ability to ‘unlock’ their data for marketing activation, and that’s ideally in a non-closed ‘stack’. However, common understanding and a desired definition for “data portability” do have a foundation in the ongoing matchrate discussion. In particular because there are further restrictions e.g. in suitable ad inventory and formats, we need to get us much of the marketer’s valuable first party data activated programmatically, as possible – for the time being, the match-rate is the best available indication here. Fully Agreed, it needs to be viewed holisticly accross all instances in the activation process.