3 Steps for Brands Handling Customer Identity Management
Customers can connect with your brand today from virtually anywhere: stores, computers, phones and call centers, to name a few. Yet it’s often tricky for brands to manage the customer data collected across multiple devices and touchpoints.
The information you gather about a customer, including name, phone number, email address, loyalty number, or IP address, can differ depending on the channel they use to reach you. Sometimes, you can easily reconcile these identifying details; but at other times, the matching process is not so clear-cut, such as when trying to square a customer who clicked on a mobile ad for a product with another who later purchased the same item from a desktop.
This is where customer identity management comes in.
Customer identity management is the process of collecting and integrating data from multiple sources to build a more complete, omnichannel view of the customer; one that helps you understand and interact with them better. This process is dependent upon capturing facts about customers, as well as drawing inferences about them. Managing customer identity is no small task. Here’s how to do it well.
Distinguish Between Facts and Inferences
If you asked marketing people what percentage of their customer knowledge was verifiable fact and what was inferred, I suspect they couldn’t tell you. Most companies don’t overtly think of their data this way. The analysts who summarize customer data may know which elements came from the customer, and which were created through combining, matching, or modeling data. However, these same details aren’t always apparent to the decision-makers using the data, which typically leads to every piece of data being treated as equally “correct.” Yet, it’s critical in good customer identity management to distinguish between the two.
Facts are the irrefutable data records of interactions and transactions. We can say with certainty that they occurred. In the vast majority of cases; we can also say with certainty that a human action caused the facts to occur, but not always.
Facts might include:
- Credit card 234234234234 was used to purchase sku XK444 at store 99 at 4:37 p.m. on May 2, 2016.
- Cookie ID 123456789 was created during session SID:ANON:mc.ai.mit.edu:NRviSpoYm7mdkYB4W2471l-01:37 and associated to IP address 188.8.131.52.
- Loyalty account 88776655 was opened on 12/15/2011 with password “cd72763” and the name “Chuck Densinger.”
Inferences, on the other hand, are conclusions drawn about the facts, and associations between the facts. They may be based on system logic, business rules, statistical probabilities, human judgment, or even customer input. Examples include:
- Chuck Densinger purchased sku XK444 because it was associated with loyalty account 88776655.
- Chuck Densinger browsed sku BQ222 because cookie ID 123456789 was associated with the browse session.
- Opens and clicks from a particular email address are from Chuck Densinger, loyalty account 88776655, because he provided that email address when he created his account (assuming he actually created the account himself).
Distinguishing between fact and inference may seem simple, but it can be tedious to go through each data element to classify them appropriately. Stick with it — there are no shortcuts here.
Establish a Strong Master Data Management Program
After data elements are clearly sorted into facts and inferences, create a process that ensures each type of data is used appropriately. Facts can be used in almost every case. With a high level of confidence in what we know, we can use them to personalize marketing and digital interactions (e.g., adding the customer’s name to an email or recommending similar products based on purchase history).
Inferences, however, must be treated more carefully. If, for example, you see two purchases from the same address with C. Densinger and Chuck Densinger, it’s possible and even likely that they were made by the same person. However, it’s also possible that Chuck has a child whose name also starts with the letter “c.” Be careful before using inferences that are revealed directly to the customer.
Master data management also covers:
- Definition of ‘Customer’: This may seem obvious, but it demands specificity. If someone visits your website and creates an account but doesn’t make a purchase, are they a customer? Once you’ve declared parameters, document the definition in clearly understood business terms and in technical terms that can be used to pull the data.
- Data Governance: Establish processes to enforce policies and standards concerning customer data management and usage. For example, who can have access to view or modify the data?
- Data Hygiene/Standardization: With data, garbage in means garbage out. Set standards to ensure that consumer data is clean and reliable, no matter who is accessing it, when they are accessing it or how they are accessing it.
- Match/De-Dupe: This is an important housekeeping step that involves linking data records belonging to a single consumer and merging duplicates.
- Data Enrichment: You don’t always have to settle for what you collect directly; adding external data to the consumer record can expand your customer understanding. This may include identifiers (name, address, email, phone) plus demographics, external purchase behavior, and media preferences.
- Persisted Master: This is the stored consumer data that is considered to be the most current and reliable. It may or may not all be in the same data location.
- Registry vs. Repository: There are two primary styles of managing the “persisted master” records. Registry (AKA, “federated”) enables different systems to “own” aspects of customer master data; whereas, repository approach establishes a central data store that is the single master for all key customer data.
- OLTP / Web / Event Services: The real-time method(s) to read, update, create, and delete access to consumer records.
- Analytic Master: Contrasted with the persisted master; the analytic master is typically not real-time, and is freed from the dictates of operational systems. It can be “omnivorous,” with respect to all interaction data, and can blend anonymous, semi-identified, and known consumer data.
If this seems like a lot to sort through, it is! But that’s what good master data management requires.
Implement Progressive Data Capture
We learn new things about our customers every day. A good customer identity management program continues to add to customer data over time and improves your ability to truly "know" which events belong to which customer. Over time, you should add the following data elements to your 360-degree view of the customer:
- web browsing behavior
- mobile app interactions
- store interactions
- third-party data appends
- call center interactions
- customer profiles
- email interactions
Digital identities and their treasure trove of valuable data help you forge critical relationships with your customers. Handle them with care, manage them properly, and you will be rewarded with customer loyalty driven by your thoughtful application of what you know (and what you infer) about your customers.
Chuck Densinger is COO of Elicit, a consultancy that helps companies transform the way they use customer and employee insight, and co-author of “Geek Nerd Suit: Breaking Down Walls, Unifying Teams and Creating Cutting Edge Customer Centricity.”