Personalization Is Not About Product, Channel or Brand

Accessing a CDP (Customer Data Platform) is a must if an organization is serious about individualized targeting and personalization. Modern marketing is about “gently nudging customers towards product and services that may interest them,” not about “bluntly pushing products and services that marketers want to sell.” That means, the first step towards intelligent targeting and personalization is getting to know your customers. Without figuring out “what they are about,” how would you pamper them properly? The key word in CDP is indeed “customer,” as personalization is not about product, channel, brand or department, but the person. To achieve comprehensive customer view, you would have to consolidate data from disparate silos. Breaking down barriers among data silos is often the main reason why companies invest resources in developing CDPs. Customer data is everywhere It is collected through online and offline stores, websites, email service providers, call centers and even machines, in forms of transaction, interaction or IOT data. Putting these data together in one place is not easy or simple; just imagine the complexity of ETL processes. Outlining the goals of CDP Connecting them through some type of ID graph is the next challenge. The main goal of CDP is to create “Customer-360” (or “360-degree view of customers”), and data of vastly different nature must be merged around “each” customer. Pretty complex endeavor for sure, and often, many service providers consider it a job-well-done, when they reach this milestone. But is the job really done? Service providers often brag about their ability to create Data Lakes. Sure, having more focused data depositories would be better than having mounds of raw data around, but I must point out that lake water is still not drinkable. Most users don’t have time or ability to plow through unrefined data, even if they may be stored in one designated place. Depending on the source, data of similar nature may look grossly different. A case in point: Let’s say you hired multiple email service vendors. How do they all maintain resolution codes for bounces, opens, clicks and conversions? Is it simply “B” for bounce, or did they add any reason codes to it? For email opens or clicks, are they kept as Yes/No flags, or timecodes for distinct actions? As you see here, putting data in one place is the easiest part. Someone must sit down and figure out how to standardize the data, and that’s the real work. And that someone should not only consider how data has been captured in the past and how they should be transformed now, but also the ways all future data is collected and consolidated. Even when outside vendors are not involved, seemingly similar data may have different values, depending on interests of different divisions within a company. For example, there can be multiple versions of product taxonomy; one for supply chain management, web display, retail and customer service. Email alert They may look different for reasons, so marketers may have to create yet another version for analytics and targeting. For instance, for website management, one may create a category called “Summer Sales.” Maybe that’s alright for one season, but should all those SKUs in that category always be labeled as “Summer Sales?” How would that help anyone figure out what products should be featured in different seasons, for each target? Speaking of “each” target, I’ve seen many cases where basic contact information isn’t consolidated in readily usable forms. For example, if someone entered multiple email addresses over time, which one should be used when contacting that person? How would you define the “best” email address? Would it be the most recently opted-in email? Or an email attached to an actual transaction? Or perhaps the email used for the latest interaction? What if there was any bounce back or opt-out? What would be your replacement rules? I am not saying that every piece of information must be in pristine condition; but you should have some idea how things should be prioritized. Unlike basic ID mapping through some commercial toolsets or service providers, creating rules for data standardization, consolidation and transformation requires deep business and domain knowledge. Almost all non-numeric information–such as PII, channels, locations, brands, products, offers, actions, reasons, interests, etc.–require meticulous attention. Often, “what’s in the data” matters more than “where they are stored.” In other words, creating golden records is not something for which you just turn on AI-based software and call it a day. Successful data transformation depends as much on business rules as on technical dexterity. Champions of CDP initiatives should never lose sight of this important development phase. Because if they don’t push the issue early on, service providers and techies may put all kinds of disparate data in one place, and just call it a day. Understanding What a Customer Data Platform Needs to Be