Retailers convinced they were hardened veterans of a demanding and unforgiving business climate — fickle shopper loyalty, fierce competitors, fast-evolving markets — faced an entirely new and previously unthinkable level of chaos during the pandemic.
The familiar stressors only intensified, and new irresistible forces exerted themselves. Supply chain disruptions proliferated. Economically distressed shoppers grew even more price sensitive while their buying patterns changed completely. Lockdowns and health concerns precipitated a dramatic shift to online channels. And the ability to rely on past years’ prices and promotions or on gut feel to navigate successfully forward has evaporated for the foreseeable future.
Fortunately, machine learning and AI-based science is well-established across the entire retail pricing lifecycle — from everyday price to promotions and markdowns. This provides even late-adopter retailers a compelling case to move toward well-established data-driven, automated, agile solutions.
[See also: Evolution of How Consumers Shop]
This technology has a proven ability to enable retailers to deliver more engaging prices and promotions that resonate with customers while delivering business goals.
Just as retailers are exploring and embracing more responsive, agile and science-based approaches to core retail processes, retail technologists are rethinking data science approaches to leverage modern capabilities and deliver solutions that can evolve at the pace of retail. A science team that is based on a culture of watching (and creating!) new innovations, and who are open to continuously evolving models and code to stay relevant, changes the game entirely.
Let’s take a look at three ways in which this modern approach benefits retailers, in contrast to thinking grounded in approaches that date back 10 years or more.
Many retail vendors trace their price science roots to academia and have an allegiance to one or two specific science approaches. As a result, they use just a few models to cover a broad range of challenges. If they try to address different phases of the pricing lifecycle, they may take a blunt approach and use, say, models best-suited to pricing and force-fit them for markdowns.
In contrast, the modern way of thinking takes a more broad-minded approach, having an array of options in both the model toolkit and the optimization toolkit to enable vendors to apply the best science for each situation.
Just as a limited toolkit results in sub-optimal science force-fitting, many traditional architectures mean that new retail scenarios will require custom code development – with the time, cost and overhead that goes with it. With a flexible architecture and a broader array of models and optimizations on hand, a modern science team can keep pace with rapidly changing shopper, competitor and market environments in retail.
Traditional one-size-fits-all approaches means that scientists must continually leverage the same components, often relying on raw compute power instead of optimized architectures to tackle a problem. This can mean increased latency, frustrated users, and expensive demands on processing capacity.
In contrast, having a broad toolset and taking a results-oriented approach enables scientists to fine-tune and optimize the application of the various science components. The result: solutions that are best able to tackle the job at hand while delivering shorter processing runtimes and lower costs.
While the promise of machine learning and AI-based science is still a gleam in the eye in many sectors, retail pricing has leveraged these approaches for many years to give retailers proven ROI, more engaging prices, enhanced competitiveness and proven ROI. Just as computer technology has continued to mature in terms of hardware and software, science and its practical applications continually evolve as well.
After well over a decade leading innovative, thoughtful and often brilliant science teams in multiple business environments and having spent many years as a faculty member teaching emerging scientists, I know firsthand how creative and curious the people are who are attracted to these fields. By working with their native strengths rather than harnessing them to old-school limitations, we can create an environment where continual discovery and innovation deliver superior outcomes for retailers worldwide — and chart a path to new innovations for the retail landscape of the future.
Geoff Pofahl, Ph.D., is head of science at DemandTec. He has more than 10 years of experience leading global data science teams to create innovative AI-based solutions leveraging deep retail pricing and promotion domain knowledge.