Making (More) Perfect Sense of Imperfect Data
If you were to ask a marketer today what their most critical priorities are, you would likely hear things like “targeting and segmentation”, “attribution”, “content and campaign optimization” and “understanding engagement across touch points”. As an industry, we find ourselves facing a tremendous amount of complexity – ultimately a function of how dramatically and rapidly the consumer has changed. Amongst all these areas that will continue to be top of mind for marketing and advertising professionals is an overriding theme: data. Data is often how we translate the “why” behind consumer behavior or campaign performance into the “how” and “what” of our marketing strategies.
Marketers almost conclusively know that all this data exists, and they also know that there’s a way to access it, thanks to the technology solutions available today. But at the same time, more than half of marketers can’t clarify, or better yet, monetize the data because their understanding of it is fragmented. According to a study done by technology provider, Signal, six in ten marketers struggle in initiatives like personalizing the customer experience and calculating ROI due to a disjointed view of their data. The reality is, you can have a lot of data platforms and advanced technology, but unless they “talk” to each other, you can’t truly unlock their value. When data is siloed, there’s no way to operate in an integrated marketing environment, and it becomes almost impossible to understand customer behavior in a way that makes it applicable to improving strategy. So the question is, how can this imperfect data be captured and cleansed in order to support the aforementioned marketing initiatives? How can we work to make it more perfect?
Audit Your Data Environment
The first step in evolving your data from just being accessible to actually being utilized is to take a hard look at the environment where all your data exists. Like I mentioned earlier, fragmented groups of data across diverse platforms will make it difficult to distinguish the data that’s adding real value from the data that isn’t.
Ask yourself these questions:
- Can I associate data to a unique customer through the creation of a persistent ID?
- Can I interpret the data based on a behavioral segment or channel?
- Can I make inferences based on demographic, geographic, or persona characteristics?
Create a Persistent ID
If your answer to the first question is “no,” it’s likely the same answer you’ll have for the following two questions. Creating a persistent ID is an essential step in unifying all your data – connecting it from multiple platforms that have different unique IDs and evolving it into a single customer identity. For example, a digital platform that uses IP address as a unique ID won’t connect directly to a platform that uses transaction number as its unique ID. By that same token, it also won’t connect to platforms that analyze demographic data or device behavior.
A persistent ID will make the association between browsing activity on a specific device type to a purchase made by, for example, an individual that is 35 years old, has a household income of $150K, and has a propensity to purchase sports and fitness products through the online channel. This once siloed data that makes limited sense within its own platform now becomes significantly more valuable through the connection by persistent ID.
Break Down the Attributes & Get Rid of Any Outliers
With a more connected view, the next step is to dissect all the data in order to understand the properties and descriptive statistics. Here is where you’ll start to make sense of the trends your data shows that ultimately become the basis for building unique customer segments, and forming strategies around them. Remember, not all of the data is going to paint a perfectly clear picture, and even more importantly, you may find that some of the trends just don’t fit. Get rid of any outliers in this step.
Build & Test Your Segments and Predictive Modeling
At this point in the process, your data has become much less imperfect, and you may even be closer to that “ah hah” moment. With the insight into behavioral patterns, channel/device preference, interactions, demographic and geographic data, you’re now in a position to build out a behavioral segmentation framework, and later predictive modeling solutions that help to increase engagement with customers, with less of the guess work on your end. You’ll want to test these models to determine their effectiveness across channels, and you’ll also want to keep an eye on how they perform over time. Customers’ behavior likely won’t remain static, and so your personas and models will need to evolve in tandem.
Taking a methodical approach like this can save you from spending thousands of dollars to restructure your data environment. A waterfall view of what was available at the start (the imperfect) to where we ultimately arrived (the more perfect) will produce a lot of insight into the overall impact on the organization and on performance across channels. If you haven’t considered this perspective yet, you’ll definitely consider it now.