6 things our client learned from their latest Text Analytics project.
You might be listening to your customers, but are you hearing what they’re really saying?
We recently analysed a large volume of customer comments for a UK company, using a combination of natural language processing, text analytics and linguistics analysis.
Here are 6 things our client (and ok, we also) learned.
- A single day can be the difference between a customer recommending you or not.
Human psychology is important.
We discovered that if a problem is fixed within seven days, then it’s fixed within a week. That’s fine.
If a problem is fixed in eight days, then that ‘week’ just became ‘weeks’.
That makes a big difference psychologically. (Similarly, if the month changes whilst someone is waiting for something to be fixed, it feels like it’s taken longer and they’ll score you more harshly.)
- Some words can act as ‘red alerts’ to help you stop an ‘issue’ becoming a ‘problem.’
We wanted to see if we could help our client predict when things might be about to go seriously wrong.
We could: we discovered two unexpected words that correlated strongly with subsequent low NPS scores.
Now that our client knows these words, when they hear them they can react quickly and give these customers extra attention before bigger problems arise.
- NPS scores can go up when your customers have a problem – it’s all about how you deal with the problem.
When we ran linguistics on the text analytics results, we heard something unusual. Lots of employees’ first names.
We were able to show that there were two key things our client’s team could do to turn a ‘problem’ into a ‘recommend’ score:
Firstly, someone on their team must take personal responsibility for fixing the situation, and make sure that the customer knew their name.
Secondly, that employee made sure that they spoke to the customer regularly about what they were doing to fix the problem.
(By the way, there’s a good story about how BA turned delays into higher approval ratings. Click here to hear it.)
- Dashboards can be useful, but only if there’s enough information.
You want a monthly dashboard? No problem.
You want a monthly dashboard for every sub-department, even though on some months there will be only one or two customer responses? Umm…are you sure? Better to save your money.
- Text Analytics by itself won’t solve your problems.
We always apply a layer of linguistics analysis over the text analytics results. This time, it showed us that “It’s okay” doesn’t equal “I’m okay”.
“It’s okay” implies you fixed the problem. “I’m okay” implies you made me happy.
You don’t just have to fix the problem to get a higher score. You have to fix the customer’s feelings.
- Text Analytics and linguistics can show you the root causes of problems in a way that quant analysis on its own can’t.
Our clients were being given lower ratings than expected in one particular region, but they didn’t know why.
We tracked the issue to a particular product from a particular supplier, solely by looking closely at what customers were saying. No one had spotted the root cause of the low NPS scores before then. Even better, the supplier has been told and the problem has now been fixed. We’re waiting for the next batch of NPS scores…
To find out more about how we use text analytics to help companies improve their customer experience, email Chris.