Yes, you can use big data to drive customer loyalty. And here's how to do it

Requires wariness of outsourcing, and your 'own secret sauce', as well as bags of common sense

While replying to tweets and hoping for the best may be many companies' approach to customer and brand loyalty, big data has truly measurable applications, as Computing learned last week.

Speaking at the Computing Big Data & Analytics Summit 2015, digital services accelerator firm Digital Catapult's Chandan Rajah quoted Gartner's "line of prescriptive analytics".

"There are two ways to get insight," he reflected: "Explicit feedback - which has discrete values - or implicit feedback. The greater value in driving an action comes from the second part - you can create an action around it.

"By analysing this you can say ‘my customer would prefer having Y on my product' and you can ask what sort of metrics you can drive. If you collect insight just for explicit input you're stuck - there's very little you can do. You can maybe increase speed and functionality, but not the quality."

Rajah explained how Digital Catapult looks "at things that don't exist now, but will in five years. But we try to find out what kind of effect we can have, then get statistical evidence. If we do, we go to the next stage".

Davy Nys, VP of EMEA and APAC at analytics firm Pentaho agreed, stating that opinions are now becoming a particular important data currency.

"The other element we're starting to see now is 'opinionism'," he told delegates.

"What we try to explain to our customers is which opinions matter. I have a few ‘opinions' about the Eurostar, but I still use it several times a week. Some people get caught up on the opinions," he emphasised, warning that just because people want to let off steam on social media, it doesn't necessarily mean their words and views are all quantifiable. "But there are algorithms."

Richard Maples, director of data and analytics at local events aggregration app firm YPlan, said that any analysis done "needs to begin with a hypothesis", in order to remove inaccuracy brought about by absorbing opinion data.

"It's only an idea, so you need to test it out and find a way of proving it. If you're just looking at a list of opinions, those are just what people say, but if you see what people got correct, that's actionable."

In terms of measuring loyalty, Maples explained how the first thing YPlan does is "take in all the data and turn it into a set of metrics about what people are doing in terms of repeat business".

"It's about trying to get people back into the app. But people are so disparate, a metric of frequency of use isn't a true metric. Number of uses a week doesn't represnt anything - it's just a number. So instead we're trying to come up with ways to segment our users that follow a curve. So it's not about actual frequency."
He also added that outsourcing such unusual analytics can be a disadvantage.

"If we don't do it ourselves initially, we don't understand it," he said.

"The benefit of outsourcing is you can scale, but if you don't understand it first, you'll just get a load of results and not understand what they mean."

Nys concurred, adding that a lot of Pentaho's customers "see the user data and internal leverage as a competitive advantage, largely on the customer experience side".

"A lot of people are wishing to use the things they know about their customers to really improve the customer experience. But try to build your own data IP first before going to buy one off the shelf - build your own secret sauce first."

Rajah concluded: "The general belief is, the more customers engage, the more they engage with the brand. But that's not necessarily true - they are now demanding a personalised experience. So segmenting a customer first doesn't work as you need to segment the behaviour. It's a slight variation on the model."