Deep learning maturing in analytics field: Teradata: Page 2 of 2
By Edwin Yapp July 18, 2018
The chief data officer
To capitalise on such trends, Brobst believes that companies today need to hire the right skill sets in order to marshal and direct such initiatives.
Commonly known as the chief data officer (CDO), companies wanting to take advantage of using AI and deep learning in their respective organisations should make investments in this area.
Asked if a traditional chief information officer (CIO) could fulfil the role of a CDO, Brobst said that it was possible but the role of a CDO had more to do with the skill set of the person rather than his or her role in the company.
“A good CDO understands the value of data but at the same time does not get caught up with the nitty-gritty details of technology,” he explained. He must not be afraid to experiment or get into technology and take calculated chances by exploiting technology.”
Quizzed as to what kind of profile this person should have, Brobst suggested that companies look at ex-consultants who have strategy experience, especially those who have run projects with technology as a driving force.
“He or she must also be able to understand how to innovate with data, manage portfolios and take risks when necessary and not be afraid to do so to reap rewards.”
Besides this, Brobst also suggested that companies take a ‘portfolio approach’ to innovation. Derived from the world of finance, a portfolio approach is a method used by investors to make investments balancing risk against performance in a phased approach.
Brobst argued a similar method could be used in today’s business world. “One could approach innovation by allocating 30% of time and resources to improving your core strengths and teaching people to do what they are already doing, better.
“Then perhaps one could allocate 60% of time and resources to trying something adjacent to your core business, and 10% to ‘moonshot’ activities – something out of this world, something you would expect to fail but learn from,” he said, noting that this would help companies embrace an innovation mindset.
Using data wisely
Last year, Digital News Asia (DNA) reported Brobst as saying that customer analytics are still lacking in banks. Many banks were still using age-old methods of reaching their customers such as telemarketing over the phone or via electronic mailers – both of which annoy many customers.
The Teradata executive then argued then that many banks today still operate on an old data paradigm, which is based on batch processing and not on real-time processing. Banks do this based on the fact that the incremental cost to their overall operations are low.
“The problem is that banks aren’t listening to what customers really want. Their financial model is broken because this old model does not account for the negative sentiments of the call or e-mail, and if banks call or e-mail too often, customers will block you or add you to their spam list.
“So, when a bank has something really useful to say, the customer won’t listen. Banks have not accounted for this cost – the cost of potentially losing the right to talk to you. Banks will have to look at the lost opportunity cost and also the possibility that the customer will churn [leave the bank for another],” he had then said.
Asked if anything has changed since last year, Brobst acknowledged that not all banks have progressed but some have become more sophisticated in using machine learning algorithms to figure out how to approach customers better with machine learning techniques rather than relying on what humans are doing, such as sending SMS or electronic mailers out to customers. This has helped them better manage the challenges they face, he said.
“When humans are involved, they rely on very simplistic cost calculations but when we look at best practices today, humans shouldn’t be involved in choosing what offers customers get.
“There is a human, say a marketing person, who defines what kind of offers a customer should get but beyond that, banks should be using machine learning to calculate who gets what offers at the appropriate time.
“This is because the machine is easily able to individualise the customer’s wants, personalise offers more to the individual’s needs,” Brobst explained, noting that this is one way machine learning has helped minimise spam e-mail to customers.