Deterministic vs. Probabilistic: A Case of Mistaken Identity
This article was originally published on Adweek.com on October 26, 2018.
If you want to market to actual people, you need to know who those people really are.
Sounds simple enough, but identity is an area where many marketers today fall short. They’re using fuzzy facsimiles of consumers based on partial profiles. That can impact their ability to understand who they’re reaching with their media and what impact that has on their overall business results.
Much of this occurs because there are a lot of misconceptions about what identity should look like. Identity is not a cookie, a device ID or an email address. Applied effectively, identity provides a 360-degree view of the consumer no matter what device they’re using, what channel they’re engaging in, or whether they’re online or offline.
If you miss out on identity, you miss out on talking to the entirety of your customer. An accurate identity is authoritative, meaning it is accurate, consolidated and up-to-date. It is persistent, in that it grows and changes as the needs and behaviors of the specific consumer change. It is omnichannel, bridging the gap between physical and digital. And finally, it is safe, providing actionable, pseudonymous information, respecting the consumer’s privacy.
So, to get identity right, you may need to rethink your strategy to ensure you’re able to do true customer-centric marketing.
Start with offline identifiers
It makes sense that many marketers look at identity as a way to recognize consumers as they move from one device to the next. But while cross-device identification is an important part of identity, it is not the whole thing.
In fact, we take the view that identity starts with actually understanding the consumer from an offline perspective — what is the person’s name, address, multiple phone numbers or multiple email addresses. Each of those is an important part of identity and only when you have a complete understanding of who the person is from this offline perspective can you begin to connect that to their multiple devices.
Be aware that many of these fractional offline identifiers come in a variety of forms. Consider this: marketers regularly refer to me as Michael Schoen, Michael A. Schoen, Michel Schoen, Michaela Schoen. All of these misspellings and errors don’t change who I am, but not accounting for them could mean you’re seeing me as different people. That’s why you need to piece them all together into an authoritative identity before you associate them with devices.
Go beyond deterministic vs. probabilistic
Another area that marketers can get caught up in is the difference between deterministic and probabilistic identifiers. They think that deterministic identity is a gold standard because it links an identifier directly to an individual. Probabilistic, because it is constructed in more generalized terms, can look more like guesswork. But that’s just not true. Each method has an important role in identity.
Let’s start by using the more straightforward, less misleading terms of explicit or implicit data.
Explicit — or deterministic — data might include offline identifiers like an email address that is associated with a cookie or a mobile ad ID. But just because an identifier is explicit doesn’t mean that it is true. For instance, my colleague might log in to my computer to check on something while we’re traveling. That’s a known signal that might associate my device with that person, but it doesn’t mean that my laptop belongs to my colleague.
Implicit data, meanwhile, comes from what consumers do during their various interactions on the web. Those signals can be used to draw conclusions that can be more accurate than seemingly unambiguous explicit data. In the example of my colleague using my computer, we might use implicit (aka probabilistic) data to determine that it’s wrong to associate my colleague to that device. The implicit data — how the device travels with me, how it is linked to the same IP address as my phone — clearly links the laptop to me.
Build accurate linkages
Using explicit and implicit data together enables you to bolster signals and associate different devices with individual people. But while you may end up getting greater scale, the cost may be less accuracy. Which is why you need to think about explicit and implicit as the methods that can check each other and deliver a more accurate customer identity graph.
For example, we see marketers today who are acquiring some sort of deterministic graph and then separately combining it with a probabilistic graph. Again, that does not guarantee scale or accuracy. In that case, it would be possible that my laptop might get thrown into my colleague’s graph. If you then add a probabilistic graph on top of that, it’s possible that it would include my colleague’s other devices in my graph as well, creating an even bigger mess.
If instead, you start with a strong initial signal at build time, you would know that my colleague’s devices travel together, separate from me and my devices and that deterministic signals triggered from device sharing are noise to be ignored.
Identity is the foundation
This kind of accurate identity can change your approach to customer-centric marketing. Marketers may already use identity in activities such as onboarding. But many don’t yet understand how that identity impacts the relative effectiveness of their campaigns. They’re still focused on media exposure data and the algorithms that determine things like multi-touch attribution. But without accurate identity, their analytics may recognize a marketing action, but they won’t actually recognize the individual.
It is important to remember, as well, that identity is not static. Marketers need to be sure they are constantly keeping identity up to date by keeping their CRM data fresh. We’ve found, for instance, that within two years, 60% of customers have had some change to their offline identity — they’ve moved, they’ve changed their phone number or email address, they’ve gotten a new job, they’ve changed their last name. If marketers are blind to that, then they won’t be able to market effectively.
It’s critical for marketers to make identity pivotal to customer-centric campaigns. After all, you can’t put people in the center of your marketing if you can’t recognize them in the first place.