The science of market segmentation and identification of customer archetypes or personas is evolving at a jaw-dropping rate. It’s another impact of Moore’s Law, and an example of how the availability of more affordablele and abundant computing power is delivering sharper digital insight into an analog world.

Segmentation involves analyzing a business’ customer population according to predefined attributes, such as age, gender, marital status, ethnicity, income, product- or service- preferences, purchase history or tendencies, and organizing the results in an orderly way. With each customer represented by its data per category - i.e. Customer A: 26 year old male, unmarried, hispanic, etc. - the database of customers is analyzed to uncover the largest groupings or clusters of attributes. Whereas previously a company might have asked, “What does my average customer look like,” it would now seek to determine the top clusters or personas in its customer population.

There are several different algorithms to conduct the analysis, but all seek to determine not a single average or distribution of customers, but rather a grouping of several average personas that together closely represent the entire population. Skillfully done, such analysis replaces anecdote-driven decision-making with data-driven decision-making, such that marketing professionals can target their messaging on the basis of the company’s customer data itself.

As a general rule, the larger and more detailed the dataset, the higher the resolution of the resulting output. Stories such as Target’s pregnancy index, already years old, point to the impressive and sometimes creepy accuracy that’s possible with basic customer attributes and purchasing history.

But when those rough customer attributes are supplemented by social media and point-of-sale data, the resulting personas can be sharpened considerably from there. Machine learning enables faster, more accurate feedback on targeted advertising campaigns, enabling companies to measure the effectiveness of their messaging to different customer personas and to refine them as they go.

As with all analytics, planning and conception are at least as important for persona analysis as the execution. While bigger is generally better, deciding which attributes to prioritize and how many personas to assemble will improve the quality and usefulness of the eventual output. While machine learning and data analytics are powerful tools, they are not yet magic bullets.

See more examples of Machine Learning in our Everyday Encounters blog series >>