Leaders from Amex, Opendoor, Cigna and Accenture talk data at Transform’s Big Bytes in AI & Data breakfast

Great applications rely on good data, just like an automobile relies on good oil. At the Big Bytes in AI & Data breakfast, presented by Accenture, panelists spoke about the importance of data, and how their organizations use it, from analyzing and normalizing, to how they ensure accuracy and reliability, and remove bias.

As the session’s host, Valerie Nygaard, product lead at Google Duplex, put it: “You can make tons of tech innovations, but so much of the time they rely on the quality of the data, that accuracy, the normalization, the processing, and the handling.”

So much of the challenge in managing data is sheer volume of it. For Accenture’s Ahmed Chakraborty, global managing director, applied intelligence North America lead, the a-ha moment came during the past six to 12 months of the pandemic as organizations were fast-tracking their digitization.

“It’s a humongous task to take 10,000 to 20,000 of data elements across the organization and manage that,” he explained. “One of the pieces of key research we did found there are only 100 to 200 data elements that are really meaningful. We call them critical data elements.”

These critical data elements have a direct co-relation to the analytics and AI that can be  done with them, and the business value you can drive on top of that. Drilling down to these most valuable data elements makes the whole effort manageable, as well as implementable, as well as high impact.

For one of their large retail consumer goods clients, they used 100-plus data elements — data per store, data per week type, data per day type, data coming from weather and so on — to build out their analytics stack, which enabled them to drive incremental revenue growth in their enterprise. And it provided huge speed to value — with that approach, they completed the project in half the time that was initially budgeted.

Data drives Opendoor’s first project, which enables homeowners to get an instant offer on their home. These valuation models operate at scale, servicing more than 90,000 customers, enabling more than $10 billion in real estate over 30 markets. Of course, the model is only as worthwhile as their data input. To ensure coverage and accuracy, Opendoor doesn’t settle for third-party data. They’ve built custom inspector apps that use a human expert to collect first-party data and then input it back into their central repository in real time. These human experts perform entity resolution when an algorithm flags conflicting data, inspect the accuracy of data, and more.

The American Express credit and fraud risk group uses credit and fraud risk models powered by machine learning, said Anjali Dewan, vice president of risk management, consumer marketing and enterprise personalization decision science. These run across the customer life cycle, starting with the opening of new accounts, assigning credit limit, all the way to fraud detection. The risk models in place are monitoring $1.2 trillion in charges annually around the world, and return 8 billion risk decisions in real time — in just milliseconds.

The human element is essential for data quality, however, said Dewan. When COVID hit last year and lockdowns came into effect, the company saw their customers’ spending behavior change virtually overnight as they began to respond to the quarantine with bulk spending and online shopping. Their machine learning algorithms which track data quality picked up these trends and started sending fraud alerts. But their data scientists recognized the behavior behind the trend, and in a matter of days they were able to update the decision science portion of their fraud engine.

“The investment we’ve made in monitoring and having our data scientists look at results from monitoring, updating them, making them better, has really helped us have our customers’ backs,” she said.

One of the challenges Mark Clare, enterprise head of data strategy and enablement at Evernorth/Cigna, has faced throughout his career is the time it takes to collect and manage data, ensure it’s high quality and governed, and then organize it to drive insights. One financial institution’s traditional data management techniques were taking months to years. Clare and his team helped them change that to just weeks with new agile, collaborative processes and visual-based discovery; the company’s global head of business discovered an eight-figure attrition risk in less than 30 minutes.

As the panel drew to a close, Dewan added that data is a vast space, and the focus on exploring new data sources internally and externally has been transformative — but quality is key.

“Having the discipline to make sure that the quality of that data stays, starting from evaluation when you put it into production, that’s key competitive advantage,” she explained. “It helps make sure that customers get the experience that you promised when you came up with the strategy and when you came up with the decision or when you came up with the model.”

Wong agreed about the importance of data quality, especially in the context of identifying a very clear objective for the business and having a measurable way of understanding how the data inputs impact the quality of your outputs.

“A lot of that goes into properly monitoring inputs and outputs,” he said. “Making sure there’s a clear mapping between the two will help accelerate the progress of adopting data and making that more usable for the company, and ultimately for your customers.”

One big takeaway for Chakraborty is that when you take a data-driven journey in the enterprise, it’s a change journey, and a big part of the change is to drive adoption.

“I call this the last-mile connection,” he said. “Literacy around data is critical. Elevating the entire acumen of your enterprise to understand data, understand what you can do with data, is so critical in the long term journey to drive adoption and the change in your culture.”