The rate of change in marketing capabilities is accelerating every day.  The barrage of technology and capabilities available to marketers seems endless. Each shiny new “technology toy” promises improved results.

Now along comes a true paradigm-shifting technology. One that marketers need to know about for its quantum shift and ROI improving potential.  That technology is machine learning.

Machine learning represents a true breakthrough technology for marketers.  Using machine learning, marketers can hyper-drive predictive analytics to yield returns that improve over time. Machine learning “learns” — fine tuning and improving campaigns as it goes.

You might have heard about machine learning as complex, expensive and difficult to implement.  Fortunately, these barriers are falling. The time is ripe for marketers to explore this approach in their organizations.

Where It All Began

Let’s be specific about what machine learning can do – enhanced predictive modeling.  Since the early days of mail-order catalogs, marketers have use predictive models to prioritize prospects for acquisition campaigns and identify customers for cross-sell.  Predictive models score prospects and customers based on the likelihood of purchase.  Marketers select a group for email and direct mail, starting from the most likely on downward.

Now there’s a better way.

What is machine learning? What do marketers need to know?

Screen Shot 2016-07-01 at 6.52.56 AM

These returns are driven by three factors:

  • Uses more variables — Traditional modeling uses 6-8 variables most commonly available to build the models; machine learning uses all the variables in the database, even if some of those variables only apply to a limited group of customers.  For example, if you have number of children in your database for only 10% of customers, traditional modeling will not use that variable but machine learning will.
  • Employs sophisticated statistical techniques — Traditional modeling uses standard regression-based modeling techniques; machine learning combines multiple sophisticated techniques to drive a more precise answer
  • Gets smarter over time — Traditional models are built once and then updated semi-annually or annually; machine learning models are “self-learning algorithms” where the model reruns frequently (daily, weekly, etc.) and becomes more accurate every time it runs, since the model will include response data from the previous week or day to make the model more accurate.

Machine learning has been around for years, but the cost in terms of dollars and time has been prohibitive.  With the decrease in cost of storage and processing, and the growth of Python, a programming language used heavily in machine learning, those barriers are breaking down.

After leveraging machine learning to improve response modeling, the next step is dynamic segmentation, where your customer groupings adjust automatically over time.  I will address that opportunity in our next post.

If you are not thinking about applying machine learning to your marketing efforts, you are falling behind.  No marketer can ignore the potential of those incremental gains in this competitive environment.

Look for our next post: Machine Learning – Where to Begin? – coming in the next few weeks as well as some case studies of marketing applications of machine learning,