Separate fads from
trends with agile
assortments & AI

In the early days of grocery retailing, assortment management involved just a handful of factors. Stores were highly localized, as was the population. People had similar backgrounds and values, often spending their entire lives in one town. So local store owners knew pretty much what shoppers wanted.

Today, retailers operate national and international chains. Their customer bases are now ethnically and economically diverse, on the move, and constantly changing their channel affinities. Grocery retailers have therefore developed increasingly sophisticated ways to match products to people. They have become pioneers in collecting and applying consumer data to merchandise mixes.

Now, the proliferation of cloud computing is enabling big data and artificial intelligence (AI) to give grocers the opportunity and power to make the assortment management and optimization process even more timely, precisely aligned to customer needs and behaviors and, ultimately, more profitable.

Solutions, powered by big data and AI, provide a Key points deeper understanding of the what, why, and how behind consumer purchasing decisions. They support a retailer’s understanding of how and when these factors change and what influences the many shifts in the market.

Assortment optimization is a process that helps retailers decide how many and which products should be offered in a category. This allows them to meet the needs of their most-valued customers and potentially attract shoppers from competitors.

In making assortment decisions, grocers compile data from point of sale (POS) systems, loyalty programs and syndicated data sources to answer questions such as, which SKUs are driving category
performance.

Yet, the tools currently used to answer these questions do not always yield fully accurate insights or deliver outcomes that deliver sustainable growth
and profitability. Nor do they prescribe optimal actions to address them.

For example, category assortment reviews are often conducted once or twice per year on a strict calendar schedule. Which can be disruptive and incur heavy operational costs. It does not factor in changes in consumer behavior that occur between reviews that can influence product choice or demand.

Traditional methods tell retailers/suppliers only about products they already know about. If inaccurate forecasting information is applied, a retailer may not order enough of a popular product and needing to use open-to-buy (OTB) dollars to replenish.

If OTB dollars are used to replenish depleted stock in a standard inline category, the incremental opportunity is lost.

Historically, assortment optimization has relied heavily on POS and syndicated data sources. This data has limitations since it does not always factor in demographic changes, preferences, social sentiment, and other factors that can provide a 360-degree view of shopper behavior.

Traditional data also looks at what customers have historically purchased and can be used as a basis for planning. But by nature, it is past tense and not validated in real time against tangible measures that reflect shifting trends.

Consumer lifestyles and behaviors are always changing. A community that was once home to many families with young children, for example, could now have more empty nesters.

This would signify far less demand for children’s snacks and lunch boxes and more demand for smaller portion food packages. But relying on historical data alone will limit a retailer’s ability to quickly realize these changes.

Retailers generally accumulate large amounts of data. But they often face time and financial constraints when it comes to consolidating them in a comprehensible manner.

According to a survey from EIQ Research (“Competing in the Age of Continuous Retail Disruption”),

and more than half of retailers lack the right skills to analyze data. With this shortage in available talent, retailers are searching for machine-first approaches that will enable them to do more with the same number of employees.

Most retailers also do not like to share data with suppliers, and vice versa, even though both parties may have mutually beneficial information. While experts extol the virtues of sharing, the debate over how to change this scenario has been raging for years.

Grocers tend to rely heavily on the knowledge and data provided by seasoned vendor executives with whom they have ongoing relationships. These vendors are undoubtedly experts in their categories. But their insights alone are not enough.

Yet, when the vendor’s data is combined with other consumer metrics and analyzed with the appropriate technology, the resulting suggestions can better meet both existing demand and future item attribute preferences of shoppers.

AI uses data mining to continuously analyze data samples in real time. Recommendations are constantly being refined and updated based on what really works. It is not necessary to wait for the next calendar review to learn that an item has experienced a temporary dip or a serious drop. Only recommendations with proven outcomes are put forward.

With machine learning built into the data process, automated triggers identify predictable patterns, peaks, or troughs and send alerts to retailers and suppliers. Retailers and suppliers do not have to make a special effort to seek out this information.

AI software, which uses NLP for dialogs with people, can “understand” and process requests in spoken English. A category manager may ask, for example, “What is the effect on profitability if we carry two rather than three brands of a product?”

AI’s ability to identify upcoming trends or attribute gaps before they significantly impact the market can allow grocers to drive a more effective and competitive private-label strategy. Retailers can then produce private-label items that will meet customer needs and support future category growth and profitability. And they can do so ahead of competitors.

AI-powered recommendations have highly predictable outcomes, leading to more accurate forecasting and inventory alignment. AI can help identify what quantities should be ordered based on an item’s historic sell-through. This can eliminate understocking or overstocking, both of which can cause revenue loss and, in the case of fresh food spoilage.

Complex mathematics inherent to AI also allow it to make accurate recommendations regarding complementary products, intricate cross-item relationships, adjacencies, and section sizes. With traditional assortment optimization, each SKU is considered independently of others, ignoring complex interdependency factors. Another benefit of AI is its ability to curate assortments to suit the needs of individual stores and customer groups at a granular level. Curation helps consumers find what they want quickly without overwhelming them with too many choices of items, brands, and package sizes. It also maximizes shelf space.

Traditional curation (sometimes called micromerchandising) focuses mainly on margins, volume, store size, location, and what customers purchased within a particular zip code. These are important criteria. But they lack AI’s algorithmic capabilities to cross-reference a wide range of data points across a myriad of consumer indices.

By applying AI, retailers can learn, for example, if shoppers are swayed by brand over price or if they decide on a price before choosing a package size. In processing this information, AI factors in demand transference data. This data tells a retailer what a shopper will accept as a substitute in a certain category or range. It’s vital for grocers looking to curate narrower, more pinpointed assortments. In fact, Gartner’s Robert Hetu recommends that retailers audit their assortment to remove “dead inventory.” AI can be an invaluable asset in applications like this.

An AI-enabled curation can even identify items that are overall poor performers but are regularly purchased by a chain’s highest-volume shoppers. Since traditional curation focuses largely on gross margin or volume, grocers can miss opportunities to help retain valuable consumers.

AI allows retailers and suppliers to continually improve category optimization and challenge the traditional approaches currently employed.

By compiling shopper sentiments in meaningful and useful ways, in addition to identifying new items, AI-driven assortment optimization helps retailers to focus on what customers want and need. AI helps companies quickly differentiate between a developing and potentially lucrative trend versus a flashy but financially risky one. The retailer can then jump on the promising new trend and avoid or limit offerings in the faddish one. AI also helps retailers and vendors better differentiate between what consumers really want versus what they say they do.

In light of AI’s benefits, a grocer’s practice of staging set, calendar-directed category reviews should be evaluated to find out if they have enough value to outweigh the disruption and operational costs they incur. AI can even be tested against traditional methods in categories that are coming up for review to see which method delivers the most net value.

By using AI in assortment planning and optimization, grocery chains could rapidly identify new items and trends before their competitors do. At the same time, they can eliminate merchandise that is no longer relevant and prepare more effectively to sell through on those lines. This would differentiate their stores, make them a destination, and give them a sharp and profitable competitive edge.

AI also helps retailers keep in sync with the everchanging needs and desires of today’s demanding consumers by staying ahead of them and curating assortments that will be future proof of category performance and market competitiveness.