In the last century, marketing has done a 180. We started with mass marketing, which ignored segment differences and sought to sell one thing with one message to as many people as possible. Today, marketers are exploiting as many differences as they can find, generating as many nuanced messages as they can manage and directing them at individual, largely digital consumers. Mass marketing still exists, but individualized marketing is King.
The name of the game is personalization, which is simply an exercise in segmenting or grouping people based on known characteristics.
But, in identifying a segment of people, who determines which characteristics are important? How do they know they’re the right ones? Can they account for change within that group – or outside of it – to ensure their target is accurate?
Most often, people are asked to make these decisions, and like all of us, these people are subject to bias and human error. While based on data, decisions around segmentation are in large part a series of educated guesses. They are assumptions about the future made by looking at the past.
One Part Data, Two Parts Decision
Imagine that you work for a regional bank, and have been asked to identify a segment of lapsing customers. You must identify customers with low or no engagement and try to re-engage them with relevant content. To do this you can look at trends across your entire customer base and decide on the specific indicators that signal a lapsing customer.
Is it 30 days with no activity? 60 days? Should the requirement be a complete lack of all engagement or if someone opened an email or logged into the website, are they still considered lapsing? Which demographics are important to consider? Do younger customers have patterns of behavior that are distinct from older generations and thereby abide by a different timeframe or activity threshold? Or does their physical location have more of an impact?
For the answer to all of these questions, you may be thinking…well, I don’t know! Certainly you could test some of your hypotheses and measure the results. But that could cost you valuable time, resources, and potentially result in lost customers. Eventually you will simply have to make a call and define hard rules for these indicators. To implement them you will create filters to hone in on this list of customers, make a wish, and deploy your message.
Enter the Machine
This dilemma is faced by marketers across industries who are trying to do the same thing. They have the data, and they are using it, but they could improve outcomes and save immense amounts of time with the help of machine learning. Instead of guessing at the answers to these questions, and then applying rules and filters to build the audience, segmentation strategies that rely on a machine learning algorithm can analyze the data and determine the right threshold for each individual person. Where one person might actually be lapsing after 30 days, another may be a loyal customer who simply doesn’t have the need to visit very often and engages only a few times a year. Furthermore, the algorithm continues to learn from the data and can adapt the segment over time to ensure accuracy. Customer behavior is always changing so static segments can become outdated quickly.
Clutch Powered Segments
Clutch Powered Segments take the guesswork out of segmentation, aiding marketers in identifying the right customers to include and optimizing the segment over time. Our powerful machine learning algorithm uses transactional data to predict future behavior of customers and is 80% more accurate at determining identifying the right customers and predicting their future actions correctly. The ROI of better targeting is clear – CPS segments result in 2X the revenue generated from traditional filter based segmentation.
Check out our data sheet on Clutch Powered Segments for more information and compelling use cases from three industries.