Are You Ready for Machine Learning?

Jeff Erhardt, CEO of

In this era of Big Data, we’re inundated with messages to use data to drive personalized and predictive experiences for customers. But for many, the gap between this data-driven vision and reality is large and growing; according to Forrester Research, most organizations are using a mere 12 percent of their data to make marketing decisions.

One approach to curing these data overload ills is machine learning, an approach to predictive analytics that allows computers to automatically learn by example and continuously adapt to the shifting sales and marketing world. Machine learning alleviates the burden of establishing fixed rules for how to react to data, extracting patterns from every individual customer and interaction in order to help you achieve the business results you want.

But how do you get started with machine learning? Are you even ready? Here are four recommendations to help you answer those questions and get your sales, marketing and support data ready for machine learning.

Don’t boil the ocean

Too many Big Data analytics projects fail by trying to solve everything at once: the 360-degree view of every possible interaction with every customer on every channel. Don’t fall into that trap. Instead, come from the philosophy that you don’t have to solve every analytics challenge right out of the gate.

Choose the right business process

Building on the mindset of not boiling the ocean, segment your business processes to identify where it makes sense to layer on machine learning technology. Start with a process that is relatively mature and stable, and where you see your employees performing repetitive tasks. Maybe it’s a process around assigning leads to salespeople or selecting dynamic content for your next email campaign.

Understand your data

Next up, take inventory to understand what data is available, where it lives and how it’s being used in that business process to make decisions. Since machine learning can work with your existing CRM, marketing automation and support systems —  and any data types, including text — there’s no need to uproot existing systems. Don’t fret if your data is scattered, or if you feel like you “don’t have enough.” Consider an agile approach; initially, you can apply machine learning using the easily accessible data and then gradually add more over time.

Be very clear about the outcome you’re striving for

Are you trying to convert more prospects? Or maybe figure out the best salesperson to close a certain deal or save an at-risk customer? The outcome is the nucleus around which machine learning operates, so the crisper the outcome, the better the value. By understanding past outcomes — good, bad, or ugly — machine learning can tie your data together in meaningful ways to drive better future outcomes for marketing and sales.

Once you understand your business process, data and outcomes, you’re ready to bring in machine learning to codify intuition —  that is, to learn the patterns of human decision making and infer the complex rules by which decisions are being made by your customer-facing employees. In turn, you’ll be able to use machine learning to scale and enhance those inherently personalized customer interaction decisions.

Predictive applications are not a black box for decision making, but a supplement to human analysis. Think about which decisions to automate and which to augment by assessing the “cost of being wrong” — the cost of an incorrect action. For example, choosing which salesperson should work a deal can be automated; the prospect likely wouldn’t notice if he engaged with the “wrong” salesperson. However, saving an at-risk customer may be better left to an augmented approach. Machine learning can help determine which customers are most likely to cancel and recommend which save technique is likely to work best, but the account manager ultimately controls how to handle the interaction.

Machine learning will play an increasing role in marketing and sales organizations looking to work smarter, not harder. By offering one more way to alleviate the burden of data analysis,
it can be the foundation for more personalized interactions and better engagement at every stage of the customer’s lifecycle.  

Jeff Erhardt is CEO of, which delivers predictive applications across the full customer lifecycle to optimize how businesses acquire, monetize, support and retain customers.