Chatting Science and Retail with Revionics' Mike Ryan

Editor's Note: With 2020 in the rearview mirror and the retail industry still adjusting to changes brought on by COVID-19, we thought it would be a good time to check in with Mike Ryan, senior vice president of science at Revionics, an Aptos Company. Read on for his assessment of the current retail environment and his perspective on the opportunities for retailers to embrace AI.

What do you see as the lasting effects of COVID-19 on retail?

MR: The pandemic hasn't changed things as much as it has accelerated necessary transformation. For most retailers, the pandemic uncovered some soft spots in omnichannel strategy and operations, exposing the areas of the business that needed the most attention. In 2020, we had a glimpse of what it would look and feel like to be caught totally unprepared, and that left an indelible impression, informing retailers' strategies moving forward.

In your mind, what matters most in 2021 for retail?

MR: What matters most in 2021 is adjusting to unified commerce and investing in the ecosystem of solutions needed to support it. The trend of more and more retail purchases moving online was greatly accelerated in 2020 with COVID-19. We've now hit that tipping point at which retailers can't operate brick-and-mortar stores and digital touchpoints in separate systems. Everything needs to work together in a single platform to support both worlds, highlighting the importance of having technology that integrates well.

What recommendations do you have for retail pricing teams?

MR: There's certainly a lot going on right now, and we're in quite interesting times. I say focus! Don't try to solve all the things all at once. Work with your technology partners and say, "These are our strategic areas of interest. How can you help?" Take deliberate and intentional steps, asking, "Which of these systems need to work together first, and for what purposes?"

I feel it's most important at this time to adopt a customer retention strategy, serving promotions to a targeted segment and playing to customer affinities – e.g., this person really values this thing, and giving them XYZ will keep them coming back. That's most important. Focus on keeping the most loyal of your customers, and then grow from there.

There's a strong appetite for AI across many retail solutions, but data problems remain. Where do you see retailers being able to embrace AI?

MR: You're right – managing data is still quite an obstacle for many retailers. I think an important prerequisite to unlocking certain strategic applications of AI is having an established customer data platform, having that as the data hub. It is important for retailers to make that adjustment.

Let's get real "science-y." What nuances of AI and deep learning do retail pricing teams need to think about as they look to the future, continuing to innovate as they go?

MR: I'd say the most important thing is not to be siloed or stuck in any one methodology while the world changes around us. An increasingly important theme in retail is that of optimizing the customer experience, but that brings new challenges, especially with data.

When you disaggregate the customer, looking at a certain consumer at a particular location, you're naturally dealing with less data than you would with a more generalized set of customers. This is a challenge for pricing and promotions.

For that reason, retailers must be able to accommodate and respond to systematic experimentation in order to be successful, putting into practice things like A/B testing and multivariate testing practices. They must also balance between neural networks, or deep networks, and the generalizability or "explainability" of AI models that aid in the human understanding of the science's recommendation.

Can you explain more about neural networks and what retailers should watch out for when it comes to ‘black box' AI?

MR: In our minds, neural network models, which are a class of black-box models, are not an acceptable paradigm for pricing and promotions decisions in general. This kind of technology gives companies the opportunity to throw all sorts of tough problems at it, and it spits out answers. The problem is that even domain experts find the results difficult to explain. If you're on a pricing team and making a recommendation, you need to have well-placed confidence in your decisions and to really understand the factors that drive any pricing action.

Transparency in our modeling, forecasting and optimization is a design choice made by our organization. I feel that transparency was really evident as we worked with retailers in 2020; our technology is built to help relieve any skepticism retailers may have in AI-powered pricing and promotions recommendations.

How do you personally keep pace with the rapidly evolving tech world?

MR: I read a variety of tech blogs on Medium. One resource I pay close attention to is DeepMind Safety Research. I love keeping up with what others in the AI and tech spaces are thinking about, what the big AI milestones are and what research is being done.

If you would like to continue the conversation with Mike about all things AI, retail pricing and how our Science teams are focused on bringing highly trusted – and transparent – AI/ML solutions to retailers, feel free to connect with him on LinkedIn.