In this episode of Simon Data’s Data Unlocked podcast, Chris Harrison discusses the role and future of identity in marketing, how identity resolution can help companies understand their customers better, our ongoing partnership with Simon Data, and more!
Summary
Chris Harrison joined Simon Data’s Data Unlocked podcast for a conversation about what happens when a brand cannot tell which of its customers is which.
He starts with ownership. Over the past decade a lot of companies handed their customer knowledge to large third parties, and now pay for access to information about their own customers. His framing is blunt: to a brand, the information about a person is the person, because that is the only way the brand knows them.
The part worth listening for is a worked example of misattributed identity walking a retailer toward an expensive strategic mistake. From there the conversation gets into how our graph actually decides which fragments belong together, and closes on why building that capability in house rarely makes sense.
What We Cover
- Misattributed identity produces confident, wrong conclusions. A retailer analyzing purchases of a trend-led clothing line found its buyers concentrated in an age bracket two decades older than the target market, a large enough signal to justify changing inventory and creative. The real explanation was adult children living at home and ordering to their parents’ address, under a second email address the retailer had never linked. Without that connection, the data looked like a demographic shift.
- Most traffic is unidentified, and most sales come from people you have already met. A typical ecommerce brand sees the large majority of revenue from repeat customers while authenticated traffic sits below a fifth of visits. The gap between those two numbers is where intent is visible and identity is not.
- A graph is nodes, edges, and a judgment about each edge. Billions of contact fragments with connections between them, where every connection carries a degree of certainty. We call that a belief vector. When confidence across a cluster is high enough, we treat that cluster as a person. Resolution is turning observed signal into a representation of someone.
- You can hold identity without holding the sensitive data. One organization wanted real-time recognition without the risk of storing personal information. We encrypted that data and set it aside, and they work from a non-reversible identifier instead, meaning one that cannot be analyzed backwards to reveal the person. They kept the recognition and shed the storage risk.
- A CDP and an identity layer solve different halves. A CDP breaks down silos inside your company. We bind the data together at the person level. One is about where the data sits, the other about who it belongs to.
- The problem is not new, only harder. Harrison reads out a slide from a presentation he gave a CPG client in 1998: identify your customers, understand what characterizes your best ones, react when engagement declines. Every line still applies, which is the point.
