As everyone working in customer loyalty knows, this industry can be confusing. This is for a variety of reasons. With greater competition, brands have to figure out ways to differentiate themselves while also maintaining a relevant value proposition. In addition, marketers have to future-proof themselves against disruptive technologies that seem to emerge every week. They also have to comply with tightening data regulations. This is a disruptive environment, to say the least.
One element that contributes to industry confusion is vague terminology. Marketers often discuss things like “machine learning,” “personalization,” or “artificial intelligence” with completely different definitions of what they are. This can lead to misunderstandings, unproductive conversations, and, if not modified, marketing goals based on misconceptions.
To address one such issue, Loyalty360 recently spoke with Michael Harrison, Managing Director at Winterberry Group, on the subject of customer data platforms (CDPs). Essentially, CDPs offer data management capabilities. Unfortunately, though, people working in the industry define the term in a variety of different ways. Forrester & Gartner define a CDP as a central customer data respository. On the other hand, companies like Zylotech, who offer a central data platform with sophisticated analytics, believe they represent the age of the sophisticated CDP.
Harrison said, “The majority of CDPs we’ve spoken to do not think of themselves as CDPs but call themselves CDPs so they get included in RFPs. This just goes to show that there’s confusion in the market. We started with the hypothesis that CDPs are data management, analytics, and applications. They do everything. However, we realized quickly that, at the core, CDPs are modern-day databases. Their capabilities are to ingest and integrate customer data.”
He continued, “They need to offer customer profile management. They need to support real-time segmentation, high-level segmentation that can fire events. Cornerstone to a CDP is that it makes data available to other systems outside the CDP. So really, CDPs should power all the application layers.”
That seems straightforward enough. CDPs manage customer data to support segmentation and to spread information into applications for other processes (personalization being one example.) However, discussion of CDPs gets complicated very quickly.
“Part of the confusion around CDPs,” Harrison said, “is that they’ve all come from a different heritage. There are very few CDPs that were built to be CDPs.” That does, indeed, make things confusing. If a CDP has been built to do something other than the work that a CDP usually does, then people won’t even think to call that thing a CDP.
Harrison continued, “CDPs store digital behavior, whether it’s curfews or device IDs, but they don’t associate that to an individual. The individual has to be authenticated first. Then they can power accuracy.” This is to say, CDPs typically aren’t used for attribution purposes.
Harrison did offer a caveat to this characterization. “What’s interesting,” he admitted, “is that CDPs are moving towards multi-touch attribution. If you’ve taken all the data and resolved all the behavior to an identity, then why wouldn’t you be able to do attribution? The hardest part about attribution is getting the data resolved for the identity. If you resolve that problem, putting attribution on the back end makes total sense.
Some CDPs also offer data visualization. “Venn diagrams, waterflows, stuff like that,” Harrison said. “Most of them don’t compete in the reporting market, though. One of them we just looked at has a full integration to Tableau, for example.”
So, while CDPs do seem to offer a variety of capabilities, they don’t offer solutions like personalization first and foremost. Marketers can use them to track customer data, but ultimately the plan is to use CDPs to feed data into other systems that enable brands to create and act on insights. A lot of marketers don’t seem to understand this, though.
Harrison offered an interesting solution. “People need to come in and say, Okay, these are what martech and adtech stacks look like and these are the purposes,” he said. “Call a CDP function what it is, and then you can say you have an ESP, an analytics hub, or an intelligence hub, which could be machine learning or just condition modeling. Then you can say you have personalization. Those are fine to have. They are applications that do application work, but the CDP does data work. Some brands that aren’t complicated, that have ecommerce sites, they probably don’t need a CDP. Vendors have caused confusion in this market by saying that, Hey, everything is a CDP, and marketers hold responsibility because they say they have a data strategy, and their data strategy is to get a CDP.”
Clearly, then, the term “CDP” is one that many brands and suppliers misunderstand. CDPs, properly speaking, aren’t the engines that drive personalization. Instead, they manage data and enable marketers to input that data into other processing systems. They may offer some limited visualization, and they do, to some extent, contribute to insight creation, but they aren’t the same thing as machine learning and artificial intelligence engines.