Using Machine Learning to Predict Customer Behaviour - DataScienceCentral.com

For a service provider, being able to anticipate its customer’s behaviour has three major benefits. It can generate customer delight, prevent customer exhaustion, and improve the company’s ROI. Let’s look at each of these benefits through three different use cases in the Customer lifecycle: Complaints Management, Customer Upsell and Customer Retention.

Generating customer delight by anticipating and pro-actively managing their complaints

A dissatisfied customer, filing a complaint is difficult to manage. He is very often passionate about his claim – whether it is justified or not – and there is sometimes little which can be done to change his perception and his opinion towards the service he initially subscribed to.

If however the company could tell precisely which customers are going to complain and when, it could avoid the management of a complaint by calling them pre-emptively to enquire about their satisfaction and offer them an incentive or a boon. This would obviously go a long way in terms of building customer delight. Furthermore, this action would definitely improve the customer’s overall experience and even increase the ROI of the relationship as very often, an preemptive outbound sales call is way cheaper than a complaint management process.

As a customer, have you ever been called by your service provider to enquire on your level of satisfaction as you wanted to file a complaint and before you even had a chance to do it? Me personally never, but I’d love to be, and I don’t despair because I know that some companies are doing it! This definitely goes a long way towards improving customer experience for sure.

Avoiding customer exhaustion by knowing when to contact them and upsell them a new product or service

Squeezing more value out of customers is in the line of sight of every service provider, especially in the telco / insurance business. There is obviously nothing wrong with that except the fact that if the customer isn’t interested to be locked in with another contract for another 12-24 months, he is going to be disgruntled by an annoying phone call from an offshore sales rep trying to push for a service he’s not interested in.

The main issue here is that the ROI of the outbound upsell campaign very often exceeds the cost of annoying your customers, with is furthermore not easily quantifiable. The consequence of a poor customer experience are however very real!

What if the company could identify prior to the campaign which of their existing customers would respond favorably to the upsell offer? This would have many benefits. The first being improving the ROI of the campagin by increasing the conversion rate, consequently reducing campaigns costs and optimizing the effort. It would also go a long way towards improving customer experience, because the company wouldn’t exhaust its customer base by targeting a limited number of customers only.

Needless to say that it also could be taken advantage of to design tailored marketing actions for specific customer segments.

Improving your ROI by optimizing your customer retention programme

Beyond the customer complaint looms the customer attrition. Being able to reduce and eventually address customer attrition is the holy grail of customer marketing. The tricky part is that once the customer identified himself as a churner, the retention cost skyrockets and very often, the company has no choice but to let him go.

The company could very well setup a retention campaign but it would incur the risk of waking up the sleeping dogs, customers who wants to churn but are too lazy to do so. It really doesn’t want to give them the opportunity to do so.

However, if the company were able to identify the potential churner beforehand, it could with a significantly lower retention cost prevent to customer from leaving the company, by surgically targeting high likelihood potential churners while leaving the sleeping dogs alone. This would increase the lifetime value of its customers, but furthermore, it would also help identify the root causes of attrition and give the company the possibility to act on improving the customer experience.

So… how is this even possible?

Until now I’ve described hypothetical situations which are win-win for both the service provider and its customers. However, their implementation would require the company to be able to read its customers’ mind. Nothing less.

This is were a supervised machine learning algorithm comes in play. The general process is the following:

You identify a target behaviour, in our case, it is a specific customer action, like filing a complaint/not filing a complaint, accepting or rejecting an upsell offer, cancelling / not cancelling a service.
You establish a list of predictors which you think might have an impact on the aforementioned behaviour. There is no hard rule here but the number of predictors can vary greatly from one model to another. It can go from a few dozens to a few hundreds just to give a ball park figure. The important thing here is that each predictor should be independent from one another. Typical predictors might be the number of customer touch point in the past year, the subscription age, etc.
You build a training set – basically you associate each past behaviour exhibited by the clients with the list of associated predictor values (the observations – for example a client who resigned his contract last month had 2 touch points and was a client since 2014.) The higher the number of observation, the better. Again, no hard rule, but a number of observation in the hundreds of thousands is quite common.
You pick an algorithm which will use the training set to create a relationship (the model) between the target behaviour and the predictors by analyzing every observation. The model can be linear with algos like linear regressions or logistic regressions, or non-linear (like tree-based algos, neural networks, etc.) The algorithm choice is ultimately based on the data scientist talent and his belief the relationship between behaviour and predictors might be. The trade off between linear and non-linear is basically the parametrization compexity vs applicable use cases trade off. Non-linear’s are more adapted in more cases, but often require heavy tweaking before yielding satisfying results.
You use a subset of your training set (the validation set on which you didn’t train your model) to test the validity of your model and make sure that the predicted outcome is not too far away from the observed outcome
You apply your model to the test set, which will predict the behaviour for customers given a set of measured predictors. The outcome can be a boolean flag (yes / no) or an occurrence probability (There is 85% likelihood the customer will churn). Furthermore, some algorithms will help you identify which of the predictors carry a strong weight in the prediction (the root causes) and which ones can be discarded.

Once you have a working model, all you need to do is to set up a process which will include the predictive solution in order to provide customer delight, prevent customer exhaustion, or improve the company’s ROI (or all three together)

A final note on the performance of such predictive models, again it can vary based on a lot of factors, but a 50% complaints volume decrease has been observed in some cases, as well as an upsell sales ratio increase by 2/3 times or an attrition drop of 15-20%.

Disclaimer

My former employer GNResearch, a fully owned subsidiary of the Teleperformance group, based in France and in Italy has developed an ROI based predictive approach which allows the companies to proactively respond to various key business needs like ‘How to anticipate and manage the dissatisfaction of the customers?’, ‘How to sell more and better to existing customers?’ and ‘How to keep the most profitable customers at an appropriate cost?’ – If you’re interested in implementing such a solution, do not hesitate to contact them directly.

Link to original post: https://logomachie.wordpress.com/2016/04/11/using-machine-learning-…