Share HBR: 'THE WRONG WAY TO REDUCE CHURN' - REALLY? WHAT BIG DATA DOESN'T TELL YOU....

In the current edition of HBR the results of a 65,000 respondent churn study with a Telco showed that if you isolate those ' at risk' because of fluctuating or over/under spending and offer them a better suited plan, overall churn increased.

The researchers hypothesize that this is because people are awoken out of their inertia and advise to further cut the data to target specific patterns.

Guess what? There are already many of these seemingly 'counterintuitive' findings around. With charities, for example, it has been shown that welcome calls and engagement increased early term churn. That's because many sign ups got sucked in and the contact reminded them they didn't really want to donate continuously.

In the Telco case the article muses that it's because plans are so hard to choose and it's almost impossible to find the right one.

But if you have even the semblance of psychological or behavioural economics understanding, you would look at it differently.

When looking at pricing and bundling, even at a superficial level, we have identified a number of needs and motivations that transcend traditional segments and render them useless when it comes to sign up and retention.

For example, the 'oversizer' prefers to buy more than they think they will ever need. The motivation is to be able to use and splurge as they want without negative consequences. JUST ENJOY!

I admit, in Telco I'm right in there. Do I care if I spend 20 or 30 bucks more a month than I could? Nope, even though our fat controller says I should.

Which brings me to the second segment.

The 'Special'. These feel that no standard plan will fit them and they will always try to tweak some element to feel they are special, whether that's data inclusion, contract term or a discount or add on. There are another 4 segments.

But, in fact, there are very few that think they need to review their spending every month and make sure they maximise it. Even in health insurance, the ability to use a lot of ancillary services translates with most into little actual use.

That's where big data often gets it wrong. And it's not the data. It's the presumption of people being rational and the service provider's service is all important to customers. Unless big data starts to incorporate psychology and behavioural economics into their models, they will only tinker around the edges.

Example 1: A client had run all kinds of modelling and experimental cells to optimise their churn reduction. The clear stance was: it can't get better. But what the models couldn't deliver was insight based, category counter intuitive messaging. By just changing 2 messages we improved the results by another 9.8% in 4.weeks.

Example 2: We've just done a specialised segmentation for a Telco that includes needs states. The presumption was, again, that price is key in the sales approach. Surprise, surprise, it wasn't. For only 2 out of 5 segments price was most important - and one of them was not a desired segment.

Example 3: Recently we worked with another service provider's win back team. Bloody tough ask. Call a customer who left and convince them right there and then to be put through to the new provider and tell them they will stay with the old one.

Previously, if our client couldn't match or better the price, the consultants just gave up and sometimes even recommended the client to go with the cheaper offer. Once we workshopped our trust building approach with the pilot team, they reported a much higher confidence and work experience. They also reported a far better customer experience and increased trust. After 4 weeks, allowing for fluctuations, the pilot team's win back performance was 10% better than that of the control team - despite having used only a smidget of the trust building opportunity.

Big data is deceptively sexy. Clear numbers. No ambiguity (but guess what, human beings are). But exactly this may just send you down the wrong burrow and make you miss out on the real improvements possible.

So, big data needs to start including other criteria and work with highly correlated factors. Clients need to make sure they don't lose deep insight and valuable understanding of customers' very own psycho logic to improve their performance.

By Prabhi Singh, Training & Licensing Lead, mext market mangement consultants