Photo: RetailWire
Should analytics drive category planning?
Recent studies indicate artificial intelligence has not yet delivered the promised time and labor savings benefits to the space and assortment optimization process.
According to Symphony RetailAI’s “Retailers’ Views of Modern Planning Tools & Solutions” study, macro space planning was cited by 72 percent of grocery retailers as critical to category planning needs, followed closely behind by assortment optimization, 67 percent; and micro space planning, 65 percent.
The study, conducted by RSR Research, found nearly 40 percent agreed assortment optimization was the most urgent system to be replaced. Thirty-one percent cited a desire to utilize cloud-based solutions for store clustering and planogram automation.
“The grocery industry has never been more complex, and the stakes are high to get category planning processes right to keep pace with change, grow sales and margins and outpace the competition,” said Cheryl Sullivan, general manager and chief product officer, Symphony RetailAI, in a statement.
IDC’s “Next-Gen Merchandising Solutions” study, which came out in January, said merchants and category managers are facing new pressures, including space constraints caused by the rise of small-format stores and the scaling of omnichannel services.
IDC’s accompanying survey found that, for 42 percent of retailers, space and assortment is their top priority for investment in merchandising technology. While 65 percent consider artificial intelligence (AI) to be essential for merchandise analytics applications, half said the primary intent of investing in AI for merchandise analytics is to drive efficiency through automation.
In a column for the Harvard Business Review, Marshall Fisher, UPS Professor of Operations and Information Management at the University of Pennsylvania’s Wharton School, wrote that techniques he helped develop with several retailers over the last few years show analytics are providing “tremendous opportunity” to improve revenues and profits, including helping identify substitutes and figuring out the right metrics to use in clustering stores.
He wrote, ”Assortment-planning processes vary greatly across retailers and product segments but have one thing in common: They rely too much on human judgment and not enough on hard data that might allow a retailer to predict how customers will react to a change in the assortment.”
Discussion Questions
DISCUSSION QUESTIONS: Are merchants still relying too much on intuition rather than analytics to guide assortment and space planning? What’s holding back progress — or is a data-first approach overrated?
Poll
Loading …BrainTrust
Lucille DeHart
Principal, MKT Marketing Services/Columbus Consulting
Ron Margulis
Managing Director, RAM Communications
The number of SKUs and complexity of identifying subtle signals in the data make merchants’ job even more difficult. Part of the challenge is in the amount of data merchants must analyze, but the other challenge is in how the data is being analyzed. SKU-level sell-through has been the key measure, but these data need to be combined with other data to truly find new insights that traditional sell-through data alone cannot provide. Additionally, merchants need to do this analysis faster than ever before to adjust to changing conditions, product availability and changing consumer buying patterns.
It’s all about timing. We are slowly emerging from the singularity of a pandemic. I think it’s a bit early to presume that AI or any kind of forecast engine is going to accurately predict demand (see Target’s inventory position).
I know I must sound like a broken record, but I think for the next several months, at least through the end of the year, we’re going to have to use a combination of gut feel and guesswork. I am not at all sure where this is all going to settle out.
Then, for sure, use analytics!
Amen! I am in the same boat and while analytics will take me so far, gut instinct will have to take me the rest of the way – especially not knowing what the future will bring. Additionally, vendors have been SKU rationalizing during the pandemic, and there is no guarantee that those vendors will go back to normal production cycles – my guess right now is that they will not.
And while the vendors are partners, they also have their own interests at heart. Only you know how to look out for your customers and your store. Even smaller independents deserve to have reliable analytics, but done in a (caveat) clear and simple format, with identified, obvious product and category upsides, so you can execute the way you need to — by gut!
I agree. For analytics to be of value the data has to be relevant. We saw what happened to its value going into the pandemic and are seeing it again as we are hopefully at or nearing its end. Until companies have enough data on the “new normal” shopping patterns, a mix of an educated gut and guesswork is the best approach.
We’d prefer category managers to actually test their educated guts. You would be shocked at how many times over the years we’ve tested a win-win solution for the category and the brand — 13% of the plans tested. Only 22% improved the category.
The issue reminds me of the debate over baseball teams: Are managers and GMs too reliant on analytics at the expense of “gut instinct,” or vice versa? Putting a roster together (and making in-game calls about pitch counts, “the shift,” and so forth) probably requires a balance between the two approaches.
Should the same balance apply to inventory optimization? The answer is yes. Relying too much on data analytics can stifle the kind of risk-taking that most retailers need to grow. On the other hand, the lack of any data-based assortment planning can make the store look like a mess.
I like your baseball analogy Dick! Makes me think of Money Ball. Personally, I think there is more of a tendency to ignore some of the data analytics and AI-based insights and rely on gut-feel. As a planning analyst, you don’t want to think you can be replaced by computers. The best approach is to combine personal insights with data insights.
Data first is the new mandatory. This is like the John Henry vs. the machine folklore — and we know how that ended (look it up). Smart merchants need to rely on data and machine learning to help guide them through the complexities of supply chain/product availability, regionality and promotional strategies. I lived through the data push back and let me tell you, the machine won every time. The true balance is to incorporate the human art with the machine science. We truly can all get along.
The mistake many are making is confounding AI with analytics. Analytics may well be critical to category management, but AI is only one of many analytical techniques and by no means a proven one.
The problem of having more data than insights is getting old, especially in retail, especially in 2022. There are tools to integrate data – sales, customer, inventory and stores – but a lack of prioritization and an abundance of resistance to change for too many retailers.
There is no silver bullet, whether AI alone, or even traditional analytics coming from the right brief. With all due respect to Google and its “sentient AI”, winning requires the right insights, the right tools and the right human(s) to interpret, be creative and test/learn (quickly!).
Further, too many merchants fail to factor in customers to merchandise planning and assortments. Understand customers, plan and allocate for them. Stores can still be a denominator, but they can’t be the only one. Customers buy.
AI should be helping retailers make better decisions and not just providing greater efficiency through automation. With the need for retailers to use every inch of space productively and really understand what is happening in their stores it is imperative that they use AI based solutions for planning and assortment; it is impossible for humans to manage this amount of data in a productive manner and make the right decisions. What retailers need is to get to the lowest level of detail, not clustering stores but working at the individual store level to get the very best results, then coordinating that merchandise planning activity with the supply chain planning, inventory and delivery scheduling and they can really show great results.
Yes, too many merchants rely on their gut rather than metrics to guide assortment decisions. Some retailers even double down on “the way we’ve always done things.” It’s 2022. It’s time to replace guesswork with data for accuracy and efficiency.
Data laggards are often constrained by budgets and overwhelmed when deciding which innovations to prioritize. Retailers that invest in analytics gain a competitive edge in pricing, promotion and assortment strategies.
The answer here depends highly on the inherent level of change that is built into any given product or category. Center aisle grocery products may not have a lot of seasonal “design” or content changes. Demand may change with the seasons, but the product itself remains the same. In apparel, the overwhelming portion of the SKUs are changing — every season, every year. This year’s data on non-seasonal basics will be abundantly useful in projecting next year. This year’s fashion data will provide little guidance on next year’s fashion. So yes, a lot of intuition will be involved. Retail involves tiered levels of seasonality, complexity and, therefore, tiered levels of predictability. The hard part is balancing analytics and intuition up and down the ladder.
The guys, and they were mostly guys, who led the merchandising departments in the 1980s, ’90s and 2000s are retiring at a fast clip right now. These seat-of-the-pants managers relied on their experience for most decisions and used data to rationalize those decisions only when needed. I still remember seeing stacks of computer printouts in the corners of buyer’s offices and asking if they were ever used to support any kind of business process. The answer was routinely no.
Now, with the advent of easy and convenient dashboards that enable merchandisers to conduct accurate “what-if” research, the corner has been turned and data analytics is ruling the roost. There is still a need for the art of merchandising, especially when it comes to big decisions relying on an acute understanding of buyer behavior, but analytics will have a role there too.
You’re so right, Ron. It is not either/or, it is AND!
Amen, Dave.
The raw data needed to drive AI-based analytics is out there. We just haven’t found an efficient way to collect it. Data is critical to the retail decision making process, but what if that data is unavailable or flawed? I have always subscribed to the GIGO (garbage in, garbage out) rule. It is critical to capture and scrub the data and understand issues outside of just movement. Factors like adjacency, price elasticity, on-hand as well as outside factors like weather, media, health consciousness, etc., drive sales. We need to capture a holistic picture at store level, aisle level, and SKU level. No easy task.
Insights from all types of data should be central to making category planning decisions. There are many information sources and analytic techniques that can help planners and merchants make better decisions.
“Artificial Intelligence” is such a misleading term, implying that big data + computers + algorithms will lead to sentient machines that will replace humans. We should think of it more as “Augmented Intelligence” in which the tools will be force multipliers for human capabilities and decision making.
I’ve had a ton of experience trying to help buyers’ assortment planning processes to improve localization and personalization at the shelf. As long as the recommendations and execution match their “industry experience,” no problem. Almost 100 percent of the time, when the data suggested something new, not so much. The key was to get the buyers and category managers to agree to a trial. They were more inclined to accept future recommendations when we could show tangible results. Unfortunately, the first time one of their leadership was in the store, they saw a planogram that they didn’t think was “on brand” because some pet item was missing or the blocking wasn’t quite right — back to square one. To really get merchandising teams to embrace data-driven decisions for store assortments, a couple of things need to happen:
This is one business problem for retailers that will continue to deserve their attention, both from a systems perspective and a process perspective. To really be successful at driving incremental sales growth, retailers need to improve these capabilities.
Analytics are important — vitally important! However there is instinct that must be applied to any category planning, assortment rationale, and shopper experience that cannot be delivered through AI or advanced analytics. I have said that if you don’t view it through “merchant eyes,” then it’s possible the “merchant dies.”
With typically over 40,000 SKUs in a store, getting simple and easy to use data is more important than ever.
A well operating grocer balances the mix of product against the greater company vision and GTM strategy. To implement, they use vendors for help, listen to the customer feedback and look at reporting. Yet reporting is often overwhelming, chaotic, taking too much time. Good analytics and tech save time, quickly showing top products and categories for upside to sales and profit (not only margin, but gross dollars $).
A willing employee with a good head and heart for the business will thrive by having clear, reliable and simple analytics available to make decisions.
Category planning and assortment planning both need “AI” or analytics. Every area of a business needs analytics to support it. As a person who owned an analytics and AI company for over 20 years, I find it interesting how over used “AI” has become. I chuckle when I hear news commentators talk about AI. It’s actually confusing the market.
It’s important for people to understand that there is a difference between reports, analytics, predictive analytics, prescriptive analytics, artificial intelligence and machine learning. It’s somewhat of a continuum, but everyone seems to call all of them AI today.
Bottom line is category planning, assortment planning, supply chain planning, sales planning, marketing planning, etc all need predictive and prescriptive analytics and that could potentially evolve into AI.