To help marketers understand and utilize this new amazing approach, this post introduces Machine Learning (ML) and addresses a few issues and ideas to consider. Each of the bullet points will be explored in more detail in subsequent writings, sometimes with more technical rigor. However, it is necessary to set the stage for those posts by setting the record straight at the outset.

  • What is Machine Learning?

The great buzz word of modern analytics today is Machine Learning; as misused as “Big Data” but still as powerful as it claims to be – when used correctly.  Machine Learning, to a marketer, is actually a stronger way to predict customer behavior across channels, time periods, programs, etc. – often showing 10-30% lift in results.

The way that Machine Learning achieves that result in marketing is threefold: using a broader range of data than traditional modeling, leveraging sophisticated modeling techniques and by establishing a model that automatically improves performance over time, without needing rebuilds.

To an analytical professional, Machine Learning is nearly indistinguishable from Frequentist/Bayesian statistics taught in graduate school. It is simply a means to an end when solving data problems. However, when ML is used correctly and the business problem being solved fits the appropriate criteria, very little can outperform ML predictive modeling methods, especially ensemble methods.

  • Do you actually need Machine Learning?

Marketers are commonly told and believe that they need to be using ML right now or they are hopelessly behind. Certain situations call for a ML approach; others do not.  The best situations for an ML approach are when you have “topped out” your results from database marketing and wish to take a new approach that can dramatically improve your results.  If you are starting out with data-driven marketing, have a new analyst or do not have a strong baseline, you may not be ready for ML yet.

  • How do you start using Machine Learning in your organization?

The most important thing to remember using Machine Learning is that you need highly qualified professionals to build the models. ML involves a level of complexity that exceeds normal model building processes, especially when using open source tools, such as Python or R.  The level of complexity becomes even greater when writing code to use real time or near-real time data. Overall, if your analysts have the skill set to build ML algorithms for prediction, explore this option. The end result will simply be far greater performance with ML.

  • Where do you go from here?

You should objectively evaluate your current analytic capabilities to identify if you have internal people who are capable of building ML. Most analytic teams lack the skills and/or the experience to build ML systems efficiently.  This can be a good time to bring in an experienced professional to get you started, help you build success in your company and enhance the skills of your internal team.

If you already have a functioning, data-driven marketing environment, look to add Machine Learning because in the end, you will wish you had that power all along.

Machine Learning can be a game changer for today’s marketer, IF it is right for your organization.  To become Machine Learning smart, check back or subscribe to our blog, for future, more in-depth discussions on this important topic.