Artificial Intelligence Pricing – A CEO’s Perspective

Artificial Intelligence Pricing – A CEO’s Perspective

Artificial Intelligence Pricing – A CEO’s Perspective
April 13, 2018 Marc Hafner - CEO and Chairman of the Board

Marc H. Hafner, Chairman & Chief Executive Officer, Revionics

Removing the Confusion

There has been a ton of confusion around what Artificial Intelligence (AI) is and isn’t, and the role Machine Learning (ML) plays. Despite the hype around the supposedly new ML and AI, rest assured it is far from being new. In fact, many of the concepts and algorithms have been around since the 1950s.

AI is the broader concept that machines should be able to carry out tasks in such a way that we could consider them smart or intelligent. Machine Learning is a current application of AI based upon an idea that machines can access data and learn for themselves without explicitly being programmed to do so. Although many techniques have been around for decades and are already heavily utilized in retail, a few factors have brought it recent fame: the abundance of data in retail today at a scale that is no longer humanly possible to mine, the inability for retailers to keep up in an environment that has become hostile and expanded beyond typical borders, and the new digital shopper whose purchase behavior has no norm and whose buying power has reached unfathomable heights.

The Good Ole Days

Long gone are the days where retailers can succeed by simply repeating the same promotions every year, only updating prices every six months, and offering promotions only through Sunday newspaper flyers and lavishly decorated end caps and store displays. In bygone times, retailers had plenty of lead time to prepare, they knew their shoppers, many by name, and shoppers’ loyalty to their favorite retailers, brands and products were unbreakable and predictable. Shoppers’ behaviors were easily influenced using established pricing concepts such as the law of supply and demand: the combination of high demand and low supply meant shoppers would easily pay higher prices to get their hands on the new item or wear the latest label.

Fast Forward to Today

Today, those pricing concepts no longer apply. A recent Forrester study commissioned by Revionics revealed that when retailers raise prices due to limited availability, 59% of shoppers said they would not purchase, wait for a better price, or shop another retailer. That is a lot of lost and delayed sales.

Today’s shoppers are very price sensitive, demand complete price transparency and are no longer loyal to a single retailer, brand or product. They shop across many channels and retailers must reach them on their terms, across a variety of channels, using a vast number of promotional vehicles (circulars, endcaps, websites, mobile, social networks, etc.). They no longer have the luxury of long lead times and competition limited to one or two other local stores. Shoppers’ preferences and price sensitivities change on a dime and as fast as the last Twitter post or “like” on Facebook, and they can select from retail options globally.

The good news is that along with the digital explosion comes massive amounts of detail on shopper behavior, very granular competitive data and a shopper who wants you to know them. The bad news is that you have merely hours to make decisions using data volumes that are no longer humanly possible to mine. Even if you could mine the data, it would be impossible to simulate all the possible options available, predict the outcome of each and choose the most optimal for your business objectives.

Machine Learning – Easy, Right?

With the current publicity around ML comes masses of players claiming expertise in the field. Suddenly retailers are surrounded by vendors slapping the buzzwords “AI” and “Machine Learning” on their sales collateral, claiming they can deliver optimal pricing and promotions using ML – often at prices well below those of established, reputable vendors.

Buyer beware! ML can be thought of as a massive toolkit of algorithms and techniques. However, the tools are only as good as those who know how to leverage them to solve real-world problems around pricing and promotions. The real world is messy and today’s mounds of data are even messier! Writing production-ready research software is hard and needs to leverage years of experience applying different techniques against massive amounts of data, business rules and constraints while giving users the ability to set the dials to drive the science against category roles, strategies and business objectives.

Given the increasingly broad availability of tools for ML and AI, it’s very easy for solution providers to leverage readily available components in their solutions and to make sweeping claims – but, it’s far more difficult to deliver robust, effective, usable solutions. Solution providers claiming ML and AI expertise and capabilities should have clear and thoughtful responses to the following questions:

  • Do the solution designers have a clear understanding of the breadth of techniques available in the ML and AI toolkits, and how did that knowledge inform their decisions of which tools to use?
  • What specific approaches are employed in their solutions, and what makes them well-suited to the problem?
  • Do they have a clear understanding of the practical challenges facing ML and AI solutions? How does their approach address these concerns?
  • Is the solution aware of underlying model quality and confidence (and corresponding decision risk)? How does the system respond or behave when information is poor, training data is sparse, or uncertainty is high? And how does the solution convey this information to the user?
  • How does the solution enable articulation of goals and objectives by the users? How does the system allow users to express goals that reflect a balance among competing concerns? Does their system help users to understand these tradeoffs?
  • How does the solution address the problem of transparency? Are recommendations easily understood and controllable through configurations and guardrails?
  • Are there real-world case studies that demonstrate measurable value delivered by the solution?

Revionics: The True Pioneer in AI Pricing

Revionics has been the pioneer in leveraging ML and AI and science to model what seems to be an infinite number of scenarios, predict outcomes and recommend optimal pricing and promotions. Although newbies to the area may try to turn around our deep roots to position us as “dated,” pure common sense will say otherwise. We have mastered the toolbox, adapting the theoretical problems in AI to a wide range of real-world problems in retail pricing. We have developed productized proprietary techniques and know how to apply them to various pricing problems, which have allowed our models to learn on production data across a range of retailers for decades, making them the most advanced, intelligent models in the retail industry. In other words, we didn’t simply give our solutions a simple voice and a name.

In an earlier Revionics-commissioned Forrester study, shoppers were asked which factors were most important when determining where to shop. Price was the # 1 factor across all retail sectors including Grocery, Apparel, DIY, and Convenience. Pricing for retailers is the heartbeat of the retail enterprise. Get it wrong, and everything else will have been done in vain. If you were about to have open-heart surgery, would you want the surgeon fresh out of school still giddy from opening up their first cadaver, full of ambition with a new shiny surgical tool but lacking in experience? Or would you want a surgeon who has performed hundreds of successful heart surgeries, removed all the risks and errors and performed optimal surgeries using all the right techniques?

Today’s retail pricing today is like heart surgery; the environment is far too complicated and unforgiving – one mistake could cost you your customers, and ultimately your very existence.

AI Machine Learning tools are only as good as those who know how to leverage them. Seek vendors with proven experience and solutions in the field. Make sure their breadth of customers spans the globe and covers the full spectrum of retail segments, from food to fashion. Double-check their customer base consists of omnichannel, pureplay e-commerce and brick and mortar. Validate that their customers get proven business results, in turn demonstrating retention rates that hover near 100% across more than a decade.

i “Demystifying Price and Promotion,” a Revionics-commissioned study conducted by Forrester Consulting, November 2017.
ii “Understanding Retail Customers’ Pricing Expectations and Tolerances,” a Revionics-commissioned study conducted by Forrester Consulting, May 2017.