Text analytics reveals what people really think of your staff

One evening, around 7.30, the Head of Customer Experience for a UK Big4 supermarket was waiting by the lifts, ready to go home. The doors opened and the CEO stepped out: head in papers as normal. He looked up, said, “Hope you’re not going to present another bloody Word Cloud at the next Board meeting.” and walked on.

The Head of Customer Experience turned round and went back into his office. His problem was one of success: he’d set up a highly efficient Voice of the Customer program and now, every month 100,000 customer comments landed on his desk. The structured (tick box) answers could be processed easily. But he knew intuitively that more insight was available from the freeform, unstructured, verbal comments. Until now, he’d been using a Word Cloud to analyse the results… until now.

He rang us. Two weeks until the next Board meeting. What could we do?

Customers want to answer the questions they wished you’d asked, not the ones you wanted answered

Word Clouds and other forms of manual-processing can’t handle the volume or velocity of comments in modern VoC programs. They’re too superficial (they look for words, not themes) and too slow (you’re halfway through constructing them and the next month’s comments arrive).

Automated methods are the only possibility for real-time, actionable insights. Even with a little time required for set up, we were able to process the 100,000 comments in near real-time. In this case, we discovered insights not previously hinted at; as soon as you let customers speak, they tell you what’s bothering them (instead of only talking about what’s bothering you).

They’re not insights until they’re actionable

We used the survey’s geographical tags, frequency questions and NPS scores to segment the customers. When we ran the text analytics software, we were able to show, for example, what was really driving the unhappy customers’ dissatisfaction with staff. For the first time, our client was able to quantify the performance of different aspects of staff interaction month-on-month.

When we dived deeper, using linguistic interpretation of the comments, something else became clear. The staff weren’t doing much wrong – it was the customer’s perception of the staff member’s skill level that was misleading the conversation. This was news to the Customer Experience team and shortly after, they kicked off a staff titling program with the HR team.

Even more valuable insights were discovered when we looked at what most appealed to those shoppers who didn’t normally shop with our client – this gave more insight in how to appeal to ‘floating voters’.

Better than we can say it

The staff perception program is now rolling through the company – and although senior management changes mean we are no longer involved with the supermarket – we know that Word Clouds are now banned for ever.