Not as Simple as “Swiping Right”
Match Rates & Their Significance in Digital Advertising
The AdTech ecosystem is fraught with different platforms that specialize in particular use-cases to assist Marketers with getting the right message in front of the right people. Despite all of the hard work and sophistication that goes into presenting a message to the right people, successfully locating or matching to that individual online remains a challenge many marketers face.
In this piece, we’ll talk a bit about what is often considered a key performance indicator when a marketer is transmitting data into online advertising platforms — Match Rates. We’ll highlight some of the more common concepts discussed when evaluating Match Rates, such as:
- Defining a Match Rate
- How Different Entities Play a Role in the Match Rate
- Deterministic vs Probabilistic Methodologies (the Accuracy vs Reach Debate)
My goal is provide foundational knowledge on how match rates impact a market’s campaign and its subsequent influence on the thousands of platforms in the ecosystem (over 3,600 today per Chief MarTec’s 2016 diagram).
Defining a Match Rate
A Match Rate is the percentage of individual users successfully translated from one entity to the data’s final entity. In our post, The “Who,” we described 1st Party and 3rd Party data that may be leveraged for ad targeting. Much of the data that we outlined came from the offline perspective, whereby an Advertiser has CRM data (1st party) or uses data from a Data Vendor (3rd party). In either case, if the data being used for targeting has Personally Identifiable Information (PII) tied to it, this can be deemed “offline” data, or data that has yet to be onboarded into the “online” world. There are some digital platforms that specialize in this offline to online translation, essentially acting as a rosetta stone between an Advertiser/Data Vendor and the Destination platform (where the Advertiser/Data Vendor is trying to push their data to for a successful “Match”).
Crossing the Chasm into Digital
When these entities send their 1st or 3rd party data to a Data Onboarding platform, their initial data set could identify individuals by parameters such as: email address, name, postal information, phone number, etc. The idea is that if a human being could look at that information and know exactly who that information represents, it could be considered PII. [For those looking to learn a bit on Data Onboarding, check out my other post “The Who”]
However, those looking to reach these individuals online need to locate/Match these parameters to an online identifier in order to send them targeted messages online (ie. JSmith@adham.com → Online ID123). This is because when our friend John Smith is browsing the internet, the webpages John visits recognize him by his online ID on that webpage.
At this stage, an Advertiser has sent a Data Onboarder offline data on an individual (“Record”), which the Onboarder has ingested and reconciled with their own internal ID. The ratio of recognized records divided by the total input records is called the “Recognition Rate.” Meaning if the Advertiser sent a data file with 10mm records and the Data Onboard could locate 6mm internal ID’s for those individuals, the Data Onboarder achieved a Recognition Rate of 60%.
But, the fun doesn’t stop here. While Data Onboarders can take on other tasks beyond this offline to online translation, the Advertiser may want the Data Onboarder to distribute the onboarded records to another AdTech platform. When the Data Onboarder distributes the data to the Destination platform, this is called the “Distribution Rate”. So if the Advertiser initially sent 10mm records to the Data Onboarder which located 6mm internal ID’s for that segment, and the Data Onboarder in turn translated these 6mm ID’s into the 4mm known ID’s it has for the Destination platform, then:
Recognition Rate = 6mm/10mm = 60%
Distribution Rate = 4mm/6mm = 66%, BUT…
…Match Rate = 4mm/10mm = 40%
Since the marketer is simply focused on how many individuals from their CRM list they can target, Match rates tend to become the primary KPI in these use-cases. However, when AdTech platforms are evaluating (to put it nicely) their connections with other platforms or trying to troubleshoot data leakage during a particular workflow between platforms, Recognition and Distribution rates can come into focus.
Further, in providing a pretty standard definition of match rates, we’ve assumed that the marketer’s goal is to target the right individuals during the campaign. However, there are some marketers who may be less focused on the exact individual, and may think it’s equally important to get their message in front of a person within a household. If you’re a marketer advertising a new family mini-van, it may matter less to you whether John Smith receives the ad specifically or if any member of the Smith household receives that message. You believe that just getting the marketing message in front of a member of the household will increase the likelihood of a purchase by the family. What we’ve just touched upon, is the debate between accuracy and reach.
Deterministic vs Probabilistic
The handful of data onboarding specialists in the Adtech space have a very real debate that they’re faced with within this niche — whether to “Match” deterministically or probabilistically. Now, within the niche, a “Match” can have numerous definitions/formulas, but the general concept is: Can the Onboarder receive PII on an individual and find a corresponding online identifier to associate with that individual?
As you might imagine, this is a difficult process regardless of whether you choose to match deterministically (meaning your have a high degree of confidence that this particular online ID can be readily associated with this individual) or probabilistically (where you either have less confidence in the match or that you simply believe one of several different online ID’s you identify can be attributed to that individual).
Particularly when you speak about probabilistic matching capabilities, you may start a discussion around the importance of “Reach” (ie the ability to capture more of the individuals in the initial input file even if that means including ID’s associated with individuals you didn’t mean to include). You may also hear about things like matching at a Household-level instead of finding an online ID that can accurately match to an individual (think about a Desktop at home that multiple family-members use).
When you hear discussions predicated on “Accuracy” as opposed to “Reach,” you’re likely taking the more deterministic perspective. Some Onboarders can argue that the central point of targeted advertising is about tailoring your message to specific audiences. If that audience is full of online ID’s that don’t map to the exact individuals you’re intending to communicate with, than what’s the point of targeted advertising in the first place?
5 Key Questions about Match Rates (courtesy of LiveRamp)
In sum, a marketer’s underlying dataset can be robust, with some identifying information on many people. However, the ability to reach those same users online is often predicated on the matching capabilities of the digital platforms the marketer is working with. Therefore, Match Rates become a central KPI when running a digital marketing campaign.