What’s the big deal about Machine Learning? It’s riding high on Gartner’s hype cycle. It’s in all the trade publications. It’s real and it’s here to stay. Whether you are in marketing, operations or finance, machine learning + data can help you do what you do better. In this blog post, I outline 7 reasons why Machine Learning is here to stay: Customer Churn, Customer Segmentation, Buyer Behavior, Asset Monitoring, Demand Forecasting, Fraud Detection and Anomaly Detection.
1. Customer Churn
We often say “You have 1,000 customers on January 1. You have 1,000 customers on December 31. How many customers did you lose?” What if you could identify the characteristics of a customer at risk of leaving you? What if you could use that information to predict, with reasonable accuracy, which of your customers might select another provider?
With your trained machine learning model, you can know which variables are the levers that need to be pulled to save those relationships. Wouldn’t it be nice if you could focus your efforts on those customers at risk while changing the business processes that caused their dissatisfaction? Yes, that’s possible today and within reach – and budget – of companies of all sizes. Here’s an article in Forbes that discusses this further.
2. Customer Segmentation
How powerful would it be if you had the ability to segment customers into, not 5 or 10 segments, but 500 or 1,000 allowing greater intimacy in marketing communications? Machine learning can be used to cluster customers in ways not immediately apparent to the human eye. Now here’s a thought: what if you could also use customer segmentation to identify which customers were your most valuable and you used THAT result in a customer churn analysis so you can focus your relationship mitigation activities on those customers who bring the most value? Boom! Mic drop.
3. Buyer Behavior
Try as I might, I cannot say it better than Kurt Marko, a Forbes Contributor, in his April 8, 2015 article entitled “Using Big Data and Machine Learning to Enrich Customer Experiences.” “Collecting, correlating and analyzing data from customer interactions across channels is the key to transforming the customer experience from nightmare to nirvana. The nexus of big data and machine learning in all its forms, including predictive analytics and even neural network deep learning, are the underpinnings of well informed, highly efficient and deeply satisfying interactions that benefit both customers and business.” Nuff said.
4. Asset Monitoring
Asset monitoring falls squarely in the IoT space. So what’s it doing here? Because of what you can accomplish when you monitor an asset and you have all that data. You can watch for anomalies, or changes in patterns that a human may not be able to discern, but would be interested in knowing. Or, and this is our fave, you can apply predictive maintenance analyses. When monitoring the sensors on a piece of equipment, machine learning tools can start identifying cause and effect relationships indicating an impending failure. Armed with this information, maintenance can be scheduled for normal maintenance windows rather than reacting to unplanned downtime, which I think we can all agree, is never fun and is often costly.
5. Demand Forecasting
Forecasting is not new. It’s the science of predicting the amount of product needed in the future using historical information and any other factors that may affect demand like seasonality, promotions, product lifecycle, etc. Machine learning enables demand forecasters to run as many forecasting models as needed to generate a more accurate forecast. So instead of forecasting for a product group, machine learning enables you to forecast for every product, in every market. We have an excellent white paper on the topic.
6. Fraud Detection
Machine learning can see things people can’t. It can spin through data, finding relationships that are anomalous and therefore, suspicious. This is the essence of fraud detection in transactions. “Hmm, Phil just bought a circular saw in Mogadishu. Phil doesn’t usually buy power tools and he’s never been to Mogadishu.” While this is intended to be a simple example, it’s not far from the truth. We’ve become accustomed to receiving text messages from our bank card companies anxious to let us know that there’s been some suspicious activity using our credit cards. If I don’t get one or two of those a month, I would wonder what’s going on. Here’s an interesting blogpost detailing how PayPal uses machine learning and humans to thwart fraud.
7. Anomaly Detection
Machine learning is very skilled at identifying when something is behaving differently than normal. Spikes, dips, level changes, anything that varies from the norm. Sometimes anomalies are incredibly important. Sometimes they’re not. Unless humans know when they occur, we’ll never know. That’s the power of anomaly detection. The interesting thing about anomaly detection is that it is a fundamental algorithm used in so many other machine learning patterns. It’s essential to predictive maintenance, customer churn, fraud detection and others.
Yes, machine learning is here to stay, apologies to any Gartner Hype Cycle. It’s already become an important weapon in too many companies’ strategic differentiation arsenal. Don’t believe me? In my experience, when Harvard Business Review starts writing about a technology, its achieved critical mass. HBR doesn’t “hype”. Don’t take my word for it. Check this out: “What Every Manager Should Know about Machine Learning”