Customer Service Surveys
- If its not one thing, its another

Alzheimers DNA molecule found? No, there are other interlinking factors as well. Customer service surveys find one issue. No, its not that one thing, its all the interlinking issues surrounding that.

One of the most daunting tasks in medical detection is the search for the cause of Alzheimer’s disease. With the breakthroughs in genetic science in the late 1990s it was thought that at last the mystery had been solved. There, miles along the DNA molecule was the culprit: a gene by the name of ApoE4. It is like a flag that turns up in most cases of people with Alzheimer’s, the condition that effectively knots up and strangles the brain cells causing an onset of dementia.

Case closed? Not exactly. Scientists have since found other genes that appear to be linked to Alzheimer’s also. A recent discovery is GAB2 which, if healthy, helps prevents the effects of ApoE4. However if GAB2 is mutant, the result is an acceleration of the knotting-up process. In other words the problem isn’t just this gene or that gene: it’s the combo.

Now let’s segue to customer service and the role of research in helping find the culprit genes that cause either very poor customer service; or truly great outcomes with loyal customers and better returns.

In customer surveys the standard analytics tend to focus on the various deliverables of your company in order to find which of these are either hurting or helping the overall customer experience. Most analysts tend to see these variables as isolated issues. Fix this. Improve that.

But deeper analytic tools are showing researchers in CRM that more often than not the success factors or failure factors are combinations of these deliverables.  Take for example that favourite of 1980s measurement: the length of time that a caller waits on the phone when they call your service number.

The truth is, most people don’t worry too much if they’re answered in two rings or four rings; and it would take a very grumpy customer to swap suppliers if occasionally they have to wait 6 rings or 10 rings.

So on the face of it the rating you get for the swiftness of phone answering may not look particularly critical.

But what if there’s a mutant GAB2 gene? Imagine you supply technology to clients and for one customer there’s a rare but serious failure in this technology. Now, of course, the phone answering is hyper-critical. A long wait could be the straw that breaks the camel’s back for this customer. It isn’t the failure. It isn’t the response time. It is the combo.

In another example from the SRD data, certain conditions interact with the status of the service deliverables. For 90 per cent of your customers, service x is unimportant and won’t make of break the relationship. However for customers coming to the end of the contract period, service x could be the tie-breaker. The customer’s status interacts with service x.

The analytic software that uncovers these hidden patterns is Neural Networking; a well established form of computer learning whereby all the variables are analysed together rather than one at a time. Analysts at SRD have fed-in customer experience surveys and have found that out of, say 50 separate questions around 20 prove particularly predictive of a happy customer or a less than delighted customer.

The mystery in these analytic investigations is that some the most potent predictors of customer satisfaction fail, by themselves, to have much a correlation with the client’s overall rating for your company.

“We were taken aback” says Duncan Stuart, who designs and analyses the SRD surveys. “Some deliverables weren’t even on the radar using conventional measures; for example when we explored the Importance/Delivery ratings – but the neural networking software put these variables near the top of the list.”

The explanation comes from the fact that neural network exercises, in which a computer trained to look for patterns and combinations, goes through literally millions of iterations of the data, are fundamentally holistic in nature. They don’t look for individual “genes” so much as combinations. Stuart nicknames these variables: ‘benign tumours.’

“This approach is allowing us to diagnose a lot more deeply,” says Neil Stewart who heads SRD. “Malcolm Gladwell cited a similar case recently in one of his articles, where an analyst armed with neural network software handily beat a group of bookies at the dog track using the same form-guide information. The software found combinations of information that the experienced bookies just didn’t see.”

The significance of this analytic difference is huge. Where standard surveys and basic analytics scratch the surface, deeper analytics will discover much more profound stories – whole oilfields – below the obvious.

Says Neil Stewart: “A restaurant doesn’t lift the customer experience just by fixing the lettuce or by dimming the lights. There’s a whole network of experiential elements. You need to understand the way things interact so you can find the combo the clicks.”