The tools and methods for data analytics are continuously advancing, but when it comes to your customers there’s more to it than just technology. ANZ Bank CTO Patrick Maes shared his top tips to make data analytics work at the recent Gartner Symposium on the Gold Coast.
Single customer view
There’s nothing more frustrating for a customer who has to re-enter data when buying a product or service.
Before you start with any kind of data mining activity on customers, first thing first is to establish a single view of the customer, Maes said.
“The same the John Bob in Singapore is the same John Bob in Taiwan. Identify the customer only once, store the data somewhere and use it.
“I wonder how many companies are spending money on big data without having that fundamental building block in place?”
Maes said he empathises with organisations that are fighting for talent in a small pool. However, it’s not exactly the kind of profession you can mould someone into or turn people with somewhat related skills into data scientists.
“You can’t just say ‘how do you become a doctor?’ You go to university and study it and after seven years you’ll become a doctor.”
He said in a recent experiment with advanced Bayesian networks, it took PhD students to really bring value out of this and understand how to utilise the models.
“This is a very complex domain and you really need a specialist to be able to use analytics in proper way.”
Network with universities, research centres and vendors to get the skills needed to carry out value-adding data analytics, he said.
“We have deep relationship with MIT in Boston, and research centres, IBM. We do a lot of research with Monash and RMIT University, and universities in Singapore, China and India - with people who are doing PhDs in analytics.”
Algorithms without heuristics is meaningless, Maes said.
“Analytical engines, or correlation engines, can prove that almost a factor of 1 – 0.9, or whatever – there is high correlation between the success of Justin Bieber and the decline of the American academy.
"Once you have found the correlation or the model, you have to implement that. That model has to be maintained because the moment you stop pushing the compile button, you can be sure that your model is already out of date. Any model that cannot be explained by humans and be maintained by humans is dead."
ANZ Bank is also training Watson to assist in making financial advice. It was the first bank in the world to use Watson, according to Maes, and has been training it for two years.
“If you can automate that know how into a machine, like Watson, then our experts can then go further and deepen their know how because they get more time to learn and to become better.”
He added that Watson would not replace human financial advisors, as he cannot see how a customer facing position could be successful without the human being involved.
“We also strongly believe technology should support people and should make people’s lives better, and not just increase profits for organisations.”
Humans are visually orientated beings and our brains are more stimulated when we see visuals of data than when we see convoluted stats, Maes said. This is key to making discoveries and insight.
“One of the experiments we did with RMIT was we took all out FY2013 expenditure data, all our accounts bill, into a 3D visualisation. You can basically walk into the data and see how much we spent on this and that, in Taiwan versus Japan, so on.
“I showed it to some of my board members, and they could find things that they were not aware of just by looking at the data, by interacting in an active way with the data instead of looking at long, two dimensional spreadsheets.
“When we look at data we make it visual, we can find patterns that the machine would not be able to find.”
Customer data analytics goes well beyond profiling as it can be used to engage them into the business, Maes said.
The bank recently did an experiment with RMIT students on building a game that simulated a travel experience of all the tourist things customers can do in that city in Asia or the A/NZ region.
The customer enters their destination, budget and time they travelling. They then enter a virtual world of their destination, with analytics that tailor certain highlights to them. For example, a customer might be more of an outdoors sort of person, so the attractions that the game makes most visible to that customer might be hiking and sport activities.
The bank also highlights in the game its deals with partner organisations such as theme parks and the consequences of travelling without travel insurance.
Once the customer has built up the maximum amount of travel experience points, they can share it on their social networks.
Maes said the game was built by two students and some members of his team in two months, and is not yet in production.