Every facet of grocery retail is becoming more personalized, a trend that is jeopardizing established operating models rooted in a one-to-many mass approach. Advanced analytics has the potential to transform grocery operations, but many retailers struggle to harness these capabilities to improve performance. Out of at least a hundred documented use cases, just a small number generate a significant impact. In our discussions with grocery executives, two emerging use cases—personalized promotions and store-specific SKU selection—reinforce the value at stake and could enable grocery retailers to truly differentiate themselves in the market.
Grocery seems to be in an ideal position to harness analytics: it boasts high purchase frequency, rich customer data, and a focus on efficiency. Yet with at least 100 documented analytics use cases in grocery, retailers can have difficulty determining which pockets of innovation matter most. In our experience with leading grocers, just a fraction of use cases represent most of the value (Exhibit 1). In this article, we examine two of the highest-potential use cases: personalized promotions and store-specific SKU selection. For both, we will examine how to unlock the full value through four elements: strategic fit, data and analytics, execution, and continuous evolution.
Leading grocers recognized long ago the need to reevaluate the true incremental impact of mass promotions. Analytics now enables complex views of mass promotions, controlling for stock-up, cannibalization, and even the halo effect that promotions have on other products. When taking these factors into account, even the best grocers can expect 10 to 15 percent of promotions to dilute sales and margins.
Personalized promotions have become more relevant and higher priority. Retailers can operate these at scale because technology has evolved, and customer touchpoints for data collection and communication (especially through e-grocery and loyalty apps) have increased in recent years. When done right, promotions can provide a substantial benefit of 4 to 8 percent sales increase and 2 to 3 percent net income and EBIT uplift.
Consumers, especially younger segments, have started to accept and even expect personalized discounts based on their shopping habits. Recognizing this trend, many grocers now offer simple personalization—either through segmented promotions (selected campaigns targeted to segments with similar preferences) or by matching predefined offers with the customers who have the highest propensity to buy. These efforts typically increase sales by less than 1 percent.
To achieve the next level, grocers should take a true “customer lifetime value” approach, in which customers are notified of the right offer at the right time, with the right message, and through the right channel.
Consumers, especially younger segments, have started to accept and even expect personalized discounts based on their shopping habits.
Personalized promotions can increase customer lifetime value by tailoring campaigns that shift customer behaviors. For example, some campaigns are geared to win customers back, others encourage more top-up sales, and others aim to increase store visit frequency. Optimizing discounts alone will not be sufficient. Instead, best-in-class players choose the best timing, the most appropriate channel, and even the most appealing communication to increase the likelihood of changing customer behavior over time. Recurring interactions generate more data, which grocery retailers can use to design increasingly relevant promotions that generate long-term customer lifetime value and loyalty.
In addition, collaboration and negotiations with consumer-packaged-goods (CPG) partners will be critical—not only to maintain overall funding levels (which requires careful tracking of granular promotion redemptions) but also to increase overall effectiveness. When done well, collaboration results in a mutually beneficial arrangement for CPGs and grocers.
Data and analytics
To support high-quality mass promotions, grocers need a robust data lake (containing data on offers, transactions, and products), an analytics environment, and executional tools. For personalized promotions, these elements need to be combined with permitted data from customer-relationship-management (CRM) and loyalty systems, including customer purchase behavior, app interactions, and profile information. Supporting mass promotions also requires navigating compliance issues around the use of personally identifiable information (PII) data: depending on the geography, consumers need to be presented with an explicit opt-in feature. Our research has revealed that consumers are becoming increasingly intentional about the types of data they share and with whom.1 The way companies handle consumer data and privacy can become a point of differentiation and even a source of competitive business advantage.
Next to data, assessing the impact of personalized promotions on customer lifetime value calls for a wider set of models. For example, there will be two models to increase store visit frequency: one model pinpoints the potential for increasing a customer’s frequency, while the other recommends the best product to drive that customer to a store. Yet another model could help in selecting the right channel or message.
To properly execute personalized promotions and gain better access to cheap channels, retailers will need to augment their analytics capabilities with appropriate technologies and a suitable operating model.
Channels: To reach digitally adept customers, grocers will need to use a wider variety of distribution channels, including point-of-sale coupons, loyalty apps, email, and SMS messages, along with third-party apps such as Instacart. Retailers can launch separate initiatives (for example, gamification) to increase reach, particularly in cheaper channels. Connecting with digital-averse customers could require collaboration with direct-marketing agencies for a more tailored set of mail offers. To finance these investments, grocers could scale back spending on traditional circulars and other mass-promotion items over time.
Technology: To run thousands of individual campaigns, grocers need to implement an appropriate marketing-technology stack and potentially make changes to the underlying infrastructure—for example, allowing traditional point-of-sale (POS) systems to print individual coupons.
Operating model: Personalized promotions create a level of complexity that strains the management capabilities of traditional merchants. To be successful, retailers must encourage collaboration among marketing, category managers, analytics, and sometimes representatives of vendors. They should also establish dedicated centers of excellence (CoEs)—for example, to design and target offers. Grocers might also need to reevaluate the decision-making process around promotions (for example, CoEs could assume decision rights for specific campaigns while category managers maintain decision rights on the overall budget).
One shortcoming of mass promotions is the inability to conduct anything beyond high-level experiments. By adopting personalized promotions, grocers can massively increase the robustness of tests (such as using A/B tests on specific customer holdouts). This approach can be used to improve statistical models, eliminate unprofitable promotion types, and create a test-and-learn mentality throughout the commercial functions of the grocer.
Store-specific SKU selection
Consumers increasingly favor stores that are nearby, a trend that spawned many small-format stores in recent years. In turn, the limited store size has made store-specific assortments a higher priority and more relevant for small-format retailers. With advances in data and analytics, retailers can now provide consumers with an assortment tailored to their unique needs. The result could be sales gains of 2 to 4 percent, presenting grocers with another substantial EBIT opportunity.
Recognizing this, many grocers take an approach based on optimizing assortment modules by store cluster (that is, one assortment module for all stores in a cluster) mostly based on price tiers (premium versus entry) and store size. However, the many dimensions of local customer preferences cannot be addressed by simple clusters.
There are three steps to achieve store-specific assortment: space allocation, SKU selection, and planogramming (Exhibit 2). Today, many grocers have already customized space allocation to local customer demand, but only a few players have achieved automated store-specific SKU selection and planograms to date. For example, in 2020, German grocer REWE announced the implementation of automated, optimized planograms to support localized assortment at the store level. Similarly, Żabka, a Polish convenience-store chain, perfected store-specific assortment for the chain’s 8,000 existing stores and all stores in development. For each address in the country, Żabka can identify the right store-specific SKU selection for the nearest location and—at extremely high confidence levels—how much sales and margin it would drive. This has allowed Żabka to grow substantially over the past few years.2
In the future, we expect many players to adopt such approaches to tailor store-level assortment to unique community tastes.
A major risk when localizing assortment is introducing additional complexity without a clear benefit. Grocers should aim to develop targeted, specific visions for how localization will generate value for their business—especially when there’s a high share of small-format stores.
For many brands, localization involves changes to distribution, so grocers need to ensure their terms with CPG companies are flexible enough to allow for regular assortment changes on the local level. Many players already have a combination of general terms and existing flexibility to accommodate store differences and varied planogram sizes.
Data and analytics
Entirely localized assortment down to the store level requires more complex models; these have been conceptually and technically defined but not yet tested at scale. Grocery retailers need a wider set of algorithms to develop sophistication across the three elements.
For space allocation, a combination of category elasticities given targets (sales or margin) and a set of business rules (such as available space and refrigeration) helps to automatically find the optimal shelf space per store and category. Store-specific SKU selection requires an understanding of customer needs, which can be gained through customer decision trees. After defining the full assortment per customer need, retailers can calculate sales potential and rank SKUs per store. Similarly, planogramming can be automated to select the best number of facings and order on the shelf.
To offer an automated SKU selection per store, grocery retailers will need to enhance operating-model and technological capabilities.
Operating model: The complexity of store-specific SKU selection can outstrip the ability of a typical merchant or category manager to manage it effectively. Currently, category managers have clear visibility for how to optimize assortment modules per store cluster. However, in a store-specific SKU selection, category managers lose sight of local demands, which limits their steering capability. Organizations will need to adapt the way analytics CoEs and category managers work together, moving from actual selection to setting guidelines.
Technology: Grocers should also pursue planogram localization, which requires them to manage a larger number of designs and versions. This will require enhancement of enterprise-resource-planning (ERP) systems and new planogram software solutions.
Store-level execution: Once a retailer settles on an assortment and creates a planogram, individual stores will need to be given directions to design the planograms in their stores. In a tight labor market, care must be taken to send modular instructions to in-store employees.
One of the most challenging aspects of localizing assortment is accurately measuring the impact. Moving to full store-specific localization will require retailers to embrace more advanced statistical techniques for gauging impact rather than relying solely on one-off tests or same-store sales comparisons. Even at the store level, A/B testing is crucial to learn and continuously improve. There are challenges to conducting these tests—including operational difficulty in setting up test-and-control stores and the additional need for statistical variability to measure results with fairly small samples—but the results are worth it.
Advanced analytics is perhaps the largest emerging source of net value creation for grocers; when done right, it can generate benefits of up to one percentage point of EBIT for the next three to five years. While there are more than 100 use cases for advanced analytics, just ten of them account for 80 percent of the value at stake. As technology continues to evolve, the next wave of use cases will be personalized promotions and store-specific SKU selection—each of which can bring more than 5 percent uplift in sales.