Businesses are turning to the customer data platform (CDP) to help them bring all of their customer data together in one place for analysis and activation purposes. One of the reasons that the CDP has taken off in recent years is that after spending years optimizing their strategies in each individual marketing channel, marketers are finally ready to turn their attention to the omnichannel customer experience. They are no longer content to view their interactions with customers in each channel in isolation.
But marketers can’t do anything to improve the omnichannel customer experience until they have the ability to recognize individuals across channels.
For that reason, the concept of identity plays an essential role in the customer experience — and therefore in any customer data platform. I’ll explore the concept of identity resolution as it relates to CDPs in this blog post.
The Unified Customer Profile
Central to any CDP is a unified customer profile. Customer data can’t simply go into a black hole that spits out lists of people meeting certain requirements. Each person, whether known or anonymous, must have a single profile that contains all data relevant to him or her from a variety of sources.
This profile should include as much data as possible to help understand the prospect/customer including purchases, browsing history, email interactions, attributes, subscriptions, loyalty membership and status, interests and preferences, browser type, location, demographics, predictive scores (like expected life-time value), and more.
And this profile should include and unify as many identifying data elements such as cookies, email address, full name, physical address, phone number, system ID, etc. as possible.
This unified profile is the foundation of any customer data analysis effort or activation campaign. A company can use it to determine whether or not each individual should qualify for certain promotions, which product or content recommendations each individual should see, which channel to use to communicate, when to send an email and which one to send, and more.
Multiple Profiles for One Person
Your goal is to create a single profile for each individual, but it’s still possible to end up with multiple profiles for the same person. When data comes from different sources, it needs to be stitched together into one single profile. This is simple when there is clear identifying information (for instance, the same email address), but not every person will clearly identify themselves in every channel for every interaction. Consider these situations:
- A visitor to a B2B site fills out a form to download a report and includes her name and email address. The company now can identify her when she is on the website. However, sometime later she uses a different browser. Her named profile still exists, but after that point, a new profile is created to capture her website activity as an anonymous visitor.
- A shopper in a retail store makes several purchases with his loyalty number in a store. The retailer has his name and other identifying information through the loyalty program, as well as a history of his in-store purchases. Later, he browses the e-commerce site numerous times but never makes a purchase or identifies himself in any way, so his named in-store profile remains separate from his anonymous website behavioral profile.
- A customer of a bank or insurance company receives an email from that firm, reads the email, and then clicks through to the website. Later she calls the call center to discuss the same topic covered in the email and that she researched on the company’s website. In this scenario, she may be seen as three different people across the company’s email, web and call center systems.
The separate profiles created in these situations all contain useful visitor/customer data, but each one doesn’t fully account for the person’s true and complete identity.
Stitching Profiles Together to Resolve Identities
CDPs need to be able to stitch profiles together when the data becomes available to do so.
For instance, let’s assume that Mary Smith has a profile in a company’s CDP that encompasses data from several different channels (store, call center, mobile app, etc.). But Mary has never identified herself on her laptop. Therefore, she has one “Mary Smith” profile and one “anonymous” profile in the CDP. The CDP needs to be able to merge both of these profiles together once she identifies herself in some way — such as clicking through from an email to the company's site while on her laptop. Once it becomes clear that the anonymous person is actually Mary Smith, both profiles need to be stitched together to create a full picture of Mary.
The same is true across any profile in the CDP. It needs to constantly monitor identifying information to determine if two or more profiles represent the same identity. This can be done through deterministic matching or probabilistic (heuristic) matching.
With deterministic matching, the CDP stitches profiles together based on clear, unique identifiers such as system ID or email address. For example, if a person visits a site multiple times but never identifies himself by making a purchase, registering for a webinar, signing up for a newsletter, etc., his profile remains anonymous. The first time he does provide his email address, the CDP recognizes all of his past sessions that took place with the same cookie, and stitches his anonymous profile together with his new identified profile.
But matching isn’t a perfect science. Oftentimes, a CDP needs to use “fuzzy” logic, called probabilistic or heuristic matching. In these situations, there is no clear identifier. Instead, the CDP makes an educated guess about profiles that likely represent the same individual based on a combination of different data such as location, behavior, etc.
One final note about identity. It’s certainly not ideal to have several profiles that reflect one individual, but it’s just as bad (if not worse) to have data from more than one person reflected in a single profile. In other words, it is essential that a CDP doesn’t incorrectly merge profiles together.
Each business has a different set of identity sources and a different tolerance for error and false positives. It’s important to be able to add your own rules to instruct the CDP of your view of the trustworthiness of different data types and decide what action to take when there is a conflict. This is very business-specific, so it’s important for the business to have control in this area.
For example, you may want to ensure your CDP merges profiles if they match on one type of field — such as email address — but not on another — such as physical address. Other businesses may want to match on email address and physical address, but prioritize the email over physical address. You may also want to value some sources more highly than others if two sources conflict. For example, the email address a person clicks through to the website from may be more likely to be accurate than the email he adds to a website form (where he may add a fake email address to try to download a piece of content, for example). Or, for example, you may regard a customer’s loyaltyID as more trustworthy than the phone number.
There are different scenarios that each business should explore to determine what logic makes sense to apply.
A principal goal with customer data platforms is to enable businesses to recognize and communicate with customers and prospects across channels. Understanding the identity of each individual is the first step to making that possible.
To learn more about Evergage as a CDP, request a demo today.