How machine learning can make customer analytics easy 

analytics

Machine learning and data science are shaking up the scientific and commercial worlds. But can they say anything new about customer analytics? Actually the potential is huge, and most companies are only just beginning to grasp at it.

In my role I have been fortunate to see how a wide range of companies are working to incorporate analytics into their decision making. The truth is that many still struggle with the basics: getting data out of silos, getting the right people in with the right skills, knowing what technology to adopt and getting to an evidence led culture. Even when investment is made the best that many seem to achieve are a few cluttered and unimaginative dashboards that few people use, falling short of the huge potential that the data presents. Then when it comes to making decisions, it’s back to the spreadsheets.

As Gartner have being saying for months now, analytics is now moving from the merely descriptive and into the realm of predictive and the prescriptive – true data driven decision making. But how is this transformation supposed to happen in practice?

The answer is that in a world of increasing informational complexity, machine learning technologies can actually help simplify the picture by making the precise drivers of business goals much easier to understand and direct.

It is hard to extract new insight customer and web analytics – even experienced analysts are are often unable to explain to the business what user analytics data says about how to improve the user experience, or grow sales. An emerging response is to use a machine learning algorithm to make sense of the vast quantities of data by using it to predict the final impact numbers that we really care about. The idea is to reduce and focus down all the data into a sharp predictable point.

Any customer success metric or goal which you can track can be modelled like this if you have enough data. Instead of building a spurious model to try to explain churn, or sales, or user reported success, the analyst of today (naturally re-trained as a data scientist) can apply machine learning to predict it far better. First, they must gather all the data you have about your customer and their journey. Second, they must train an algorithm to predict the one key metric that you actually do care about. Third, they use the model to help decode cause and effect, predict the impact of changes and prioritise actions: in other words, to become truly data driven.

An enhanced white box model, such a decision tree, can be applied in this way with the great advantage of transparency. If you can look under the bonnet of the algorithm to understand how the prediction works, then you can precisely understand the link between your user data and real success.

Yes the good news is that end really is sight to manually wading through mountains of data, and puzzling over noisy dashboards. We have the technology to let the data decide – so let’s use it!

Published in
1 Comment
  1. Dan 6 months ago

    Awesome post Gabe! I need your customer focused analytics applied to the machine generated logs of cyber adversaries, to learn how to defend against them. 😉

Leave a reply

Log in with your credentials

or    

Forgot your details?

Create Account