Category Archives: Data Analytics

Predict and improve, before customers churn

It has been a while since AI (artificial intelligence) and ML (machine learning) have been part of the discussions and conversations around how technology can help companies engage with customers, and ultimately earn their business by providing them with a better experience.

The more cocky software companies shout about how the AI and ML capabilities of their technology platforms “will completely transform” our businesses, and make us successful. Other companies, maybe more realistic, say their technology platforms “can enable transformation” and deliver improvements.

But if I’m honest, it always feels that they are exaggerating and over-blowing their own capabilities, and that is probably why many of us still struggle to see how, in reality and in practice, things work and could be applied to our case, and in our business.

One of the problems I see is people try to “run before learning how to walk”. Businesses must be at a specific level of maturity and readiness to adopt certain things, and my preferred approach is always to start with the basics, and make sure it’s done right, before “embarking in bigger adventures”.

And it is also extremely important to understand, that data is the key ingredient. No technology platform or capability (including AI or ML) will deliver any outcome without data. And if you have data, there are very powerful things you can do first, and that you are probably not doing, before trying to get robots to run your business.

An example: Predictive Analysis. Technology platforms with this capability allow you to perform statistical analysis and data mining, using current and historical data, to make predictions about future behaviors. They use ML and predictive modelling to find patterns in that data, and identify risks or opportunities.

Predictive analytics could be used by commercial industries, but also by organisations that serve citizens, students or patients, to determine their behaviour, predict future engagements (purchases or interactions), or even guess if they are about to stop engaging with you.

A use case: Customer Churn. Use technology to help you predict if your customers are in risk of leaving you, or stop buying from you. Understand when they are likely to do that. And, even more powerful, why they are about to do that – which could allow you to amend, correct or enhance things before they do!

If you are going to attempt this, the first thing you should do is define, what “churn” means to your business, as it could have different definitions. A few examples are:

  • A customer cancels a subscription
  • A customer hasn’t logged-in to the website
  • A customer hasn’t purchased over a period of time (e.g. 1 year)
  • A customer has reduced their purchases (e.g. by 50%)
  • A customer has stopped engaging in a community / forum

Once you have that defined, you need to gather a significant (maybe a few thousands of records) and relevant data set (maybe the last 12 to 18 months), which includes customer, operational and experience data, and where each and every record is tagged with “churn” or “no churn”.

You could then feed your technology platform, with this data set, and allow it to build a model, which will help you (with more or less accuracy, depending on the data set you used) predict your customer’s behaviour.

Curious about which technology allows you to do this kind of thing, or how to go about implementing this?… At Capventis, the company I work for, we have helped and enabled a few clients, using technologies like Qualtrics, Qlik or Alteryx.