Mastercard is planning to embed AI-driven data analytics in the day-to-day workflow of its retail and banking customers in China, to improve the quality and efficiency of data analytics and boost returns from this technology.

A research conducted with 2,000 executives found that only 20 percent of them were getting adequate returns on the data analytics they did.

The executives gave four reasons for the surprising outcome of the research, which was jointly conducted by Mastercard and Harvard Business Review earlier this year.

"First of all, they said today's analytics is happening in silos, meaning various parts of the company are running their own analytics, and tend to produce conflicting results sometimes," said Dimitrios Dosis, president of Mastercard Advisors, during a recent interview in Beijing.

"Second, there is a big time lag between the moment you need the data and the moment you get them. Sometimes it can take weeks. Third, data analytics is not really embedded in the workflow. When people need it to make decisions, they are not getting it. And fourth, they said sometimes you need a PhD degree to understand the software and the results, which means it is not really intuitive."

The fact that data analytics is not embedded in the day-to-day workflow is one of the primary concerns of Dosis who heads Mastercard Advisors.

Offering information, consulting and implementation services to merchants and financial institutions worldwide, this unit of Mastercard helps customers cleanse and understand the data they have, including anonymized and aggregated transaction data from Mastercard, to derive recommendations for customers based on data insights and advanced analytics.

Before fully rolling out the recommendations and executing them, consulting teams from Mastercard Advisors test the recommendations through the application of a test-and-learn technology.

"What we do is identifying a concrete opportunity based on our data, specifying the targeted segments where this opportunity primarily exists and then identifying the offer, and testing and executing it. This is a classical end-to-end service we provide for many banks, including Chinese banks," Dosis said.

Right now the company is developing a technology for this end-to-end service so that data analytics will become an effective part of the day-to-day work process. That means people do not need to do specific analytics while it is happening in the background.

"Imagine that for a cards manager of a bank, when she comes in the morning, instead of her logging in and running analytics, she gets a message on her device that says, 'Looking at the data from last week, we believe you have an untapped opportunity in the mass affluent segment.'

"Automated recommendation engine provides her the right offers for the right audience and asks, 'Would you like to test it?' She says yes. Six weeks later, she gets the results, chooses the best campaign and rolls it out. The analytics is happening in the background, and she is just there to make decisions. This is the technology that is going to come next," Dosis said.

So far, deriving recommendations has been a manual process, with consultants looking at the data regularly.

Companies have a lot of data and customers would like to interact with them, but the data are not cleansed. As data cleansing takes a lot of time, artificial intelligence could be applied in the process, Dosis said.

"Normally, it took us 80 hours to analyze the data and come up with recommendations. By applying artificial intelligence and having a more automated recommendation engine, we have been able to reduce this to 10 hours," he said.