"Data-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 Chris O’Hara, vice president of strategic accounts at Krux.
Marketers comparing data management platforms have probably asked a vendor about match rates. Unfortunately, many don’t understand what they are asking for. There is simply not much plainspoken dialogue available about the topic.
Match rates are a key factor in deciding how well a vendor can provide cross-device identity mapping. Marketers are starting to request match rate numbers as a method of validation and comparison among ad tech platforms in the same way they wanted click-through rates from ad networks a few years ago. Back in the good old network days, click-through rates were the ultimate arbiter of vendor selection, leading many to game the system.
Will match rates become the de facto metric for choosing DMPs? I think this is a dangerous idea, and I’ll explain why.
As a consumer, I probably carry about 12 different user IDs, including Chrome and Mozilla cookies, IDFAs for my Apple phone and tablets, a Roku ID and a few hashed email IDs. Marketers looking to achieve true 1:1 marketing must reconcile all of those identities to a single universal consumer ID (UID).
The first, most important issue to solve before any “matching” takes place is a vendor’s ability to match people to the devices and browsers associated with them. Assuming the vendor has nailed this tricky cross-device problem, it now must match that UID against the places where the consumer can be found. This ability is the vendor’s “match rate.”
So, what’s the number? Herein lies the problem. Match rates are really hard to determine and change all the time. To figure a vendor’s real ability to match user identity is, marketers can ask two basic questions.
What Am I Matching?
What are marketers asking a vendor to match? There are two types of matches to consider. One is a vendor’s ability to match offline data to cookie IDs, which is called onboarding. The second is matching one set of cookie IDs to another and matching multiple devices to a single user.
During onboarding, offline personally identifiable information (PII) identities, such as an email, are matched with cookies. It’s widely accepted that this method matches about 40% of users in the online space. That seems pretty low, but cookies are a highly volatile form of identity, prone to frequent deletion and dependent upon a broad network of third parties to fire match pixels on behalf of the onboarder to constantly identify users.
Over time, a strong correlation between consumers’ offline IDs and their website visitation habits, combined with rigor around the collection and normalization of identity data, can yield much higher offline-to-online match results, but it takes effort. Beware the vendor that claims it can match more than 40% of emails to an active cookie ID from the get-go.
For cookie-to-cookie user mapping, the ability to match users across platforms has a lot to do with how frequently a vendor fires match pixels. This happens when one platform (a DMP) calls the other platform (the DSP) and asks, “Hey, do you know this user?” That is a one-way match. It’s even better when the latter platform fires a match pixel back: “Yes, but do you know this guy?” This creates a two-way identity match. Large data platforms will ask partners to fire multiple match pixels to make sure they are keeping up with all IDs in their ecosystem.
For example, a DMP with a big publisher client that sees most of the US population fires a match pixel for several DSPs at the same time. Every user visiting that big publisher’s site would get that publisher’s DMP master ID matched with the three separate DSP IDs.
In this scenario where there is a high degree of frequency in match pixel fires, match rates in the 70% range are still considered excellent. Beware vendor claims of 90% match rates in the cookie space. This type of matching is also a complex process and involves many parties and counterparties; numbers that high aren’t typical nor are they all that realistic.
What Populations Are You Matching?
Consider a marketer that’s gathered a mess of cookie IDs through first-party web visitors and wants to match them against a bunch of cookie IDs in a popular DSP. Vendors saying they have a 90%-plus match rate in such situations should be a red flag. Many of those online IDs are Safari IDs – not cookies – and cannot be matched. That eliminates a good 20% of matches right off the bat. Also, many of those cookies are expired and no longer matchable, which adds maybe another 20% to the equation. Taken together, about 40% of possible matches are eliminated, making a 60% match rate overall pretty good.
Lots of vendors are actually talking about a matchable population of users, subtracting from the total those users, similar to those above, that are unmatchable. Some may do this transparently: “Hey, we can’t do anything with these cookies over here.” Others may be less transparent, relying on opacity and a vague reference to matchable population so they can claim a higher match rate, airbrushing over the fact that it’s actually against only a subset of your cookies.
When a DMP fires match pixels all day, several times a day, with a favored DSP, the match rate at any given time may indeed be 90% to 100%. But that’s when measuring against a matchable population and not the whole. With fractions and percentages, the numerators and denominators matter. Always ask the follow-on questions to understand the basis for the comparisons being made.
As a side note, some might wonder whether popular DMP/DSP combo platforms offer higher match rates. Such tie-ups often tout “lossless integration,” since both the DMP and DSP carry a single architecture and unified user identity. The short answer is yes, but two independent DMP and DSP platforms, closely aligned and with an effective synch frequency, can deliver identical match rates.
Keep Asking Questions
Marketers are obsessing over match rates right now, and they should be. There is also an awful lot of “FUD” (fear, uncertainty and doubt) and, let’s be frank, a lot of BS being tossed around in terms of numbers. So, keep asking questions.
What kind of cross-device graph does a vendor support? Without the fundamental ability to understand the one-to-many relationship between people, browsers and devices, the “match rate” number is largely irrelevant.
In terms of populations, what numbers is a vendor matching? Onboarding (matching offline IDs to cookies) and cookie matching (mapping different cookie IDs in a match table) are two different problem sets. Marketers should approach each on its own terms, cognizant of the different dynamics at play.
Marketers should ask vendors how and what they are matching. How do the numerator and denominator relate to each other? How does this specific match rate percentage correspond to the total universe of consumers they want to reach?
Perhaps most importantly, marketers should bring a healthy dose of paranoia to the table. Never trust a number without an explanation. When a vendor offers an impressive match rate of 94.5% without asking a ton of questions, which should give them pause.
And simpler still? Just ask for a match test. The proof is in the pudding.