![Analytics imperial college](/business-school/sites/default/files/styles/focalpoint_5x3_960/public/2019-08/Untitled%20design%20%2863%29.jpg?h=27a93da2&itok=O-KV95HX)
Published
The number of channels for interacting with customers is proliferating. Staying on top of thousands of tweets, likes, referrals and reviews brings added complexity to customer care. Yet, it also brings opportunity, and with the right tools, the power to predict customer sentiment.
Text mining is at the forefront of such data science techniques, an approach that is changing service on two fronts: first, aggregating channels to monitor product performance and contact quality and secondly, gathering information to build next generation service, which could be anything from keeping track of all possible service channels to building responsive ‘bots’ that answer your most frequently asked questions.
So how is this technology having an impact on business in practice? If we take two examples, one large telecommunications business and the other a start-up food delivery company, their customer service problems are very different. The start-up challenge is about staying agile and responsive to customer needs – in this case, tracking events across the customer journey, such as how they react to product packaging, recipes or the delivery service. It is possible to monitor these by enabling a comment box that collects qualitative input, which when combined with all social media interactions in a central database means all comments can be bucketed into established categories. The second step is to then analyse this data by category in order to record sentiment – usually by building an AI or machine-learning classifier to automate the process. Finally, to turn all of this data into actionable insights, automation tools can raise a yellow flag if a customer is dissatisfied or a red flag if the company is about to lose the customer.
On the other hand, for the telecommunications business, with millions of interactions each month, the challenge relates to monitoring the customer service portal. While this relies on similar analysis, the categorisation is much broader than the start-up across a wider set of smaller services such as top-ups or data services. Similarly, once monitoring is online, flags can be raised for technical support, for instance, if an unusual number of queries start to crop-up relating to a specific issue, the business can check in real-time what issues need to be fixed straight away, pre-empting customer problems and effectively reducing the pipeline of customer support from three weeks to a matter of hours.