Companies need to move past using analytics and artificial intelligence to merely inform business decisions and allow the technologies to actually make decisions on their own across all functional areas, according to the director of data science at Samsung.
In his keynote at the Big Data Toronto conference on Tuesday, Matthew Fitz suggested enterprises are failing to fully capitalize on the huge gains in storage, insight generation and predictive modeling that have emerged in the last few years. He urged organizations to look beyond investing in the latest technology to essentially do the same things they’ve always done, which he described as marginal improvements over existing processes.
“We have to get smarter about building data platforms to do optimal decision-making,” he said. This includes “mundane decisions that machines can make far better than humans.”
Samsung already knows which of its products people own, for example, how they use them, how they’re interacting with them and comments they make about the company on social media. Theoretically, that should offer ways to make predictions about what kind of marketing content might be most relevant, but in practice he said the barrage of data can become overwhelming to even the sharpest of technical gurus in an enterprise.
“We have loads of data, but that doesn’t guarantee it will always be used properly to provide the greatest customer experience,” he said, adding “Business units are not always fully equipped to embrace today’s technology.”
Samsung has begun dealing with this issue by developing what Fritz described as a “prescriptive engine” that takes its 360-degree customer profiles along with data from other line-of-business applications and applies business rules and objectives the organization determines in advance. This lets the technology set thresholds for how much to spend on customer targeting in a given marketing campaign, for example, or ensures cross-channel consistency in what Samsung says in its marketing content so that customers can trust what’s said no matter where they hear it.
Prescriptive engines are developed using a range of algorithms that include linear programming, multi-arm bandit, Thompson sampling and gradient descent. Most sales and marketing execs may not be familiar with these algorithms, but Fritz said they represent building blocks that are within reach of most organizations.
Since taking a more prescriptive approach, Samsung has seen a near immediate 3x lift in several of its key performance indicators (KPIs), Fritz said.
Although Fritz’s team focuses on the consumer side of Samsung, he said the techniques he and his team have developed would apply in B2B scenarios as well. Using prescriptive engines, for instance, the company can automatically determine which kinds of products or service offerings it should be driving a particular customer towards.
“This took a bit of a mental shift at Samsung,” he admitted, in part because of how it has forced the organization to question some of its assumptions.
“The mantra at Samsung is, if you’re an early adopter you must be getting the (latest) phone every year. We believed that and wanted to support that, but that’s not how it came out,” he said, adding that a deeper analysis showed such customers were simply more disciplined in the way they choose their smartphones. “We didn’t even want to share it with the business, we felt so stupid.”
Of course, with any kind of automation there are always fears of job loss, but Fritz insisted that enabling deep, contextually-aware frameworks based on big data will organizations to become more creative in how they develop strategy and tactics. Perhaps most importantly, it will keep them on their toes in looking for the most accurate answers to questions.
“We continue to run businesses as though we know the status quo,” he said. “Taking a step back, taking a deep breath and having a machine like this take a look can show some really high lift.”