To drive marketing results, you need more than just data. Here's why and what it is that's missing.
If you know us, you know we talk a lot about data. After all, CDP stands for Customer Data Platform, right? And so data is foundational to what we do.
That’s all true, but it’s also only part of the story.
Because the truth is that data by itself doesn’t do much for marketers. Just having a list of email subscribers doesn’t really tell you what they need or when you should contact them. Having a database of phone numbers doesn’t tell you who’s likely to convert. Knowing when Customer X last visited your website doesn’t really move the marketing needle in and of itself.
Which is why when we talk about choosing a CDP, we can’t do it without talking about data science.
We recently launched a whole white paper on the topic, and you can get your free copy here. But today we thought we’d pull from what our data scientists shared there to give you some insights into how data becomes real customer insights.
If data is the information we gather about our customers, insights are the next step. They’re what happens when a data scientist (or built-in data science in a best-in-class CDP) interprets that data to help us truly understand our customers.
The data might give us a list of customers who renewed their subscriptions. An insight would tell us that the average customer has a 60% chance or renewing, and customers who get a 10% off coupon have a 70% chance of renewing.
The data might tell us that Angel Jones purchased a kitten hoodie. An insight would tell us that customers who purchase kitten hoodies are 2x as likely to buy kitten tutus.
The data might tell us 100,000 people watched our recent kitten video. An insight would tell us that customers with an affinity for our kitten care video series are 50% more likely to make a purchase than the average customer.
The data is (obviously) foundational to getting these insights. But the real value for marketers lies not in the piles of data, but in the actionable insights they provide.
Knowing that coupons drive a 10% uptick in subscriptions may push us to offer more coupons. Knowing that Angel Jones is highly likely to buy a kitten tutu means we can target her with ads for said tutus. Knowing that kitten video-watchers are likely purchasers means we can target them across our ad platforms. Insights, in other words, drive action. And action drives results.
So, we get it. Data needs insights in order to really move the marketing needle.
But how do we get those insights?
The answer is in a data science task called feature engineering. It’s the process of taking raw data and categorizing it into useful variables (known as features) for your data science models.
What does that mean in practical terms? Well, let’s say we want to predict if someone might open an email. What do we need to know to predict that?
Answers might include things like:
:: Have they opened emails in the past?
:: Have they opened emails from us recently (and how do we define recently)?
:: Do they have an affinity for the subject of the email? Do they often read blog posts or watch videos on that subject?
:: Have they opened email on that subject or a similar subject in the past?
The data that answers these questions is our features. It’s what we’ll feed into our data science model in order to predict our marketing outcome (in this case, “will this person open this email?”).
Notice that it’s more complicated than just tossing raw data into a machine and watching an insight pop out the other side. Feature engineering is the art part of data science. It’s where experienced marketers help the machine know which data it should analyze, which data is likely to be important.
Studies say this step in the data science process is where data scientists spend most of their time. And because there’s an art to the process, it's also where a lot of data platforms fall down on the job. They’ve got their algorithms. They know how to make data models. But if their data isn’t feature engineered, the accuracy of those models drops drastically.
If customer insights are your goal (and they should be), there are three core things you need (and should be asking about when evaluating vendors).
The first is the data itself. Without that, you don’t have a foundation for your insights.
The second is data science models (for more on those, download our free guide). They’re how the machine filters and understands your data.
The third is feature engineering—which is how the machine knows which data matters.
These three together help you understand customers, predict what they want, need, and will respond to, and make smart marketing decisions.
Need more data-driven customer insights? We’d love to help. Feature engineering is built into Lytics tools like Audience Discovery, Behavioral Scoring, Content Affinity, and Optimization. Which is why most of our clients see big results fast—usually within 60 days.
Get in touch with a Lytics expert