Rebuilding macro space planning on a modern analytics platform

QuantSpark replaced a leading UK supermarket group's ageing Excel VBA space-planning models with a bespoke Python-based analytics platform. That platform is now the backbone of macro space decisions across the retailer's supermarket and convenience estate. The source case study reports no single headline number here; the outcome is structural rather than a percentage or a pound figure. A process that used to take a planning team weeks to run on food and non-food ranges alike is now repeatable, transparent, and fast enough to keep pace with live scenario planning.

The problem. The retailer's macro space decisions, how much bay space to allocate to which category, in which store, at which time of year, ran on large Excel VBA models. These models collected bay-space and sales data across food and non-food ranges in both supermarket and convenience formats, and used historic sales curves to work out where space should move. Built up over years, the models had become slow to run, difficult to maintain, and hard for anyone outside their original authors to audit. Some individual analyses took weeks to complete, which meant the planning team could not iterate at the speed the business needed.

This is a familiar pattern in large retail estates. Spreadsheet tools that start as a tactical fix for a specific planning problem accumulate hidden logic over successive owners. Eventually the model becomes a bottleneck rather than an accelerant: too slow to rerun often, too opaque to explain confidently to a category director, and too fragile to hand to a new analyst without substantial retraining.

The approach. QuantSpark worked alongside the retailer's own data science and engineering teams through a four-stage transformation rather than a single rip-and-replace. First, the existing Excel tools were optimised in place, improving their speed, interface and the accuracy of the recommendations they produced, giving planners a better tool immediately while deeper work continued underneath. Second, the modelling logic itself was refined to reflect store-specific customer behaviour, rather than applying one uniform curve across a diverse estate of supermarket and convenience formats. Third, QuantSpark built a Python proof-of-concept that calculated space recommendations from location-specific shopping missions, customer need states and sales curves, replacing spreadsheet formulae with logic that could be tested, versioned and explained. Fourth, that proof-of-concept was developed into a bespoke analytics platform supporting the full macro space workflow, shortening the path from scenario planning through to implementation.

How it flows today. The workflow moves from the legacy Excel estate, where bay-space and sales data were originally collected, through refined store-behaviour logic, into the Python model that generates recommendations from missions, need states and sales curves, and out through the bespoke platform into scenario planning and implementation. Each stage of that chain replaced a manual, spreadsheet-bound step with a modelled, auditable one.

What it runs on. The starting point was Excel with VBA macros, the retailer's long-standing space-planning tool. The rebuilt recommendation engine runs in Python, chosen for the transparency and testability a spreadsheet cannot offer at this scale. The production system is a bespoke analytics platform, purpose-built for the macro space workflow rather than a repurposed off-the-shelf tool. None of these are named proprietary products in the source material; they are described here only in categorical terms.

The value delivered. The bespoke platform is now the backbone of macro space decisions across the retailer's estate, spanning both supermarket and convenience formats. It substantially increases the internal team's ability to make strategic and operational space changes, and it shortens analysis that previously took weeks. Moving the underlying logic out of opaque VBA and into a testable Python model also addresses the maintainability and audit problem the client originally raised, since a documented, versioned model is inherently easier for a category or finance director to interrogate than a decade of nested spreadsheet formulae.

Because the source reports capability and speed gains qualitatively rather than as a single quantified metric, there is no pound-saving or percentage figure to decompose into an ROI calculation here. Where a retailer wants to build that business case for a similar programme, the generic method is to measure planner hours consumed per analysis cycle before and after, multiply by the loaded cost of that time and the number of cycles run per year, and add any incremental sales uplift captured from faster, more responsive space allocation. None of those inputs are available in this case, so no such figure is presented as fact.

This case illustrates a pattern QuantSpark sees often in retail: the constraint is rarely the absence of data, it is a legacy tool that has outgrown its own architecture. Replacing the model, not just the interface, is what turns a weeks-long manual exercise into a live planning capability.