Planogram Optimisation at a European Grocery Chain — 6.4% Like-for-Like Sales Lift Across 1,200 Stores
Demand Forecaster + UX Insight Engine + ML Model Deployer reshaping per-store planograms around the actual demand patterns each store serves.
The challenge
The chain — a European grocery retailer operating approximately 1,200 stores across multiple country markets, with a portfolio spanning urban-convenience, suburban-supermarket and rural-superstore formats — was operating a planogram approach that the chain's category-management leadership considered structurally suboptimal. Planograms (the per-store shelf-layout-and-product-positioning specifications) were maintained centrally with country-level variants and limited per-store tuning. The result was that individual stores were running planograms calibrated to the country-level average rather than to the specific demand pattern the store served — and the structural variance in demand patterns across the chain's diverse store base meant the central-average planogram was suboptimal for most stores.
The opportunity was clear in the data the chain already had: stores with similar formats and similar catchment demographics consistently showed different per-SKU velocity patterns, suggesting that per-store planogram tuning would produce material sales lift. The challenge was operational: maintaining 1,200 distinct planograms manually was not feasible with the chain's category-management workforce, and previous attempts at per-store tuning had stalled on the engineering complexity of producing-and-distributing 1,200 distinct planograms.
The constraints were specific. The chain's planogram-execution workflow was deeply integrated with the chain's store-operations workflow (shelf-stocking, ordering, in-store merchandising) and could not be disrupted. The GDPR compliance posture applied to the customer-data inputs the platform would use. The category-management leadership wanted the platform to augment rather than replace the category-managers' judgement on the strategic planogram decisions.
The approach
MindMap deployed a planogram-optimisation platform composed of Demand Forecaster (Df) as the per-store-per-SKU velocity-prediction engine, UX Insight Engine (Ux) repurposed for in-store-shopper-behaviour analysis, ML Model Deployer (Ml) as the model-management layer for the per-store models, and Customer 360 (C3) for the catchment-demographics-and-loyalty-data integration.
Phase one was the per-store velocity-prediction build. The previous planogram approach had used country-average velocity data. The new approach uses per-store-per-SKU velocity prediction, with features including the store's historical velocity, the catchment demographics, the loyalty-data-derived basket-pattern features, the seasonal patterns, and the cross-SKU substitution patterns within the category.
Phase two was the planogram-optimisation build. For each store and each category, the platform generates a recommended planogram considering the per-SKU velocity prediction, the per-SKU shelf-space requirements, the category-management strategic priorities (specific SKUs the chain wants to grow regardless of pure-velocity ranking), the cross-category category-management constraints, and the planogram-execution constraints (the practical limits on how often planograms can change without disrupting store operations).
Phase three was the category-manager workflow. Category managers see the platform's recommendations alongside the existing central planograms and the per-store performance analytics, and approve, modify or override the recommendations. The platform's recommendations are advisory rather than directive — the category-manager's strategic judgement remains the system of record.
Phase four was the planogram-execution integration. Approved planograms flow into the chain's existing planogram-execution workflow (the shelf-stocking instructions to store associates, the ordering recommendations to the central supply-chain, the in-store merchandising guidance) via the chain's existing planogram-distribution platform.
The pre-built building blocks
Rather than commission a ground-up build, the engagement leaned on MindMap's pre-built accelerator library — production-tested components that compress what would otherwise be a six-to-nine-month build into weeks.
Demand Forecaster
Per-store-per-SKU velocity-prediction model
UX Insight Engine
In-store-shopper-behaviour analysis
ML Model Deployer
Per-store model-fleet management
Customer 360
Catchment-demographics and loyalty-data integration
The architecture
The platform runs on the chain's Azure tenant in the EU region with full GDPR compliance. The customer-data inputs (loyalty data, basket-pattern data, demographic-overlay data) are processed with the chain's GDPR-required consent-and-purpose-limitation framework.
Demand Forecaster's per-store-per-SKU velocity model is a gradient-boosted-tree ensemble with per-store fine-tuning. The base model captures the cross-store velocity patterns; the per-store fine-tuning captures the store-specific deviations from the base pattern. The model is trained nightly on the rolling 13-week window of per-store sales data and refreshed weekly with the longer-context retraining.
Customer 360's catchment-demographics-and-loyalty-data integration provides the per-store demographic and basket-pattern features. The integration uses the chain's existing loyalty-data warehouse (with GDPR-compliant access controls) and the third-party demographic-overlay data the chain licenses.
ML Model Deployer manages the per-store model fleet. The chain has approximately 240 distinct planogram variants (the 1,200 stores cluster into roughly 240 distinct planogram profiles, with similar stores sharing a planogram variant). Each variant has its own model, with the platform managing the model-lifecycle (training, validation, promotion, retirement) across the fleet.
The UX Insight Engine repurposing handles the in-store-shopper-behaviour analysis where the chain has in-store camera-and-sensor coverage (currently approximately 280 stores). The behaviour data feeds back into the planogram-optimisation, with the shopper-flow and shelf-engagement patterns informing the planogram recommendations.
Integration with the chain's planogram-execution platform is via the platform's standard inbound API, with the approved per-store planograms flowing into the existing execution workflow.
The numbers behind the story
Like-for-like sales have risen 6.4% across the 1,200-store estate on a rolling thirteen-week basis (controlling for category-mix shifts and macro-economic conditions). The lift is most pronounced in the categories with the highest per-store demand-pattern variance — fresh categories, beverages, snacking — where the per-store tuning has the largest absolute effect.
Stockout-event rate has dropped 31%. The better per-SKU velocity prediction has produced both better shelf-positioning (more space allocated to the SKUs that genuinely need it per store) and better ordering recommendations (the per-store ordering quantities now reflect per-store velocity rather than the country-average).
Category-manager productivity has materially improved. The platform's recommendations have eliminated much of the per-store planogram-research work that the category-managers had previously been unable to perform at scale. The category-managers' time has been redirected to the strategic category-management decisions and the supplier-engagement work that the previous workload had crowded out.
An unexpected outcome: the platform's per-store demand-pattern visibility has surfaced specific store-category combinations that consistently underperform the platform's predictions. The chain's operations team has used the insights to identify specific in-store-execution issues (positioning errors, stocking issues, in-store merchandising gaps) that the previous analytics had not surfaced.
Strategic implication: the platform has become the chain's primary basis for testing assortment and category-strategy hypotheses. Specific assortment experiments can be deployed to specific store subsets with the platform's recommendation engine, with the results measured at the store-and-SKU level and the learnings incorporated into the next-cycle category strategy.
“Per-store planogram tuning at twelve hundred-store scale had been operationally infeasible with our category-management workforce. MindMap delivered the per-store tuning with the category-manager-workflow integration that has made it sustainable, producing a six-point-four per cent like-for-like sales lift. The platform has changed how we think about category management.”— Chief Merchandising Officer· European Grocery Chain
Why MindMap was chosen
The chain had evaluated two global retail-analytics vendors. Both had per-store velocity-prediction capabilities but limited integration with the chain's planogram-execution workflow and limited category-manager-workflow support — the recommendations were technically sound but operationally inert.
MindMap's accelerator-composition approach — bringing Demand Forecaster, UX Insight Engine, ML Model Deployer and Customer 360 together with the category-manager-workflow integration and the planogram-execution integration — was the structural differentiator.
Our embedded grocery-retail expertise on the delivery team (two former category-management heads from peer European grocery chains and a former planogram-execution lead) was the third factor. The chain's CMO felt that the team understood the operational reality of planogram execution in grocery retail, not just the modelling.
Related deployments
Pan-African FMCG Forecasting
Demand Forecaster delivered SKU-market forecasts across 18 African countries — cutting stockouts 22% while reducing overall inventory 14%.
Global CPG Supplier Intelligence
Supplier Benchmarker + Anomaly Detector delivered continuous risk visibility across 12,000 Tier-1 suppliers — early warning on 84% of subsequent disruptions.
Pan-African Content Generation
Text Generator delivered 40x content throughput across 6 African languages and 18 markets — same brand voice, same compliance posture, structural cost reduction.
Want an outcome like this?
Start with a 2-week AI Readiness Sprint. We deliver a prioritised use-case backlog and business case grounded in what's actually buildable with our accelerator library.