Shell Treats the Forecourt as a Grocery Operation
Shell confirmed on July 22 that it has selected RELEX Solutions to run AI-driven forecasting, replenishment, and store operations across its UK Mobility and Convenience network. That network is larger than most people assume: more than 550 convenience stores trading under Shell Select, Little Waitrose, and Co-op banners. The plan spans fresh optimization, promotions and events planning, weather-based forecasting, fresh markdowns, and a mobile replenishment tool that lets store teams generate and adjust order proposals before they confirm. This is the toolkit of a serious grocer, and Shell is applying it to sites people still think of as petrol stations with a fridge.
The framing matters for anyone running a mixed estate. Convenience formats attached to fuel, transit, or logistics hubs have quietly become a growth channel, and they carry the hardest planning problem in retail. Small footprints, high fresh mix, and volatile footfall punish generic replenishment logic. Shell is signaling that it wants those sites planned with the same rigor as a supermarket, and that it would rather license that capability than keep improvising with spreadsheets and category-manager intuition.
Why Convenience Retail Breaks Generic Demand Planning
A 550-store convenience network is a forecasting nightmare precisely because each site behaves differently. A forecourt on a motorway sells a different basket at 7am than a city-center Little Waitrose sells at 6pm, and both swing hard on weather, local events, and promotions. Weekly average models smooth all of that into uselessness. RELEX earns its keep by forecasting at the store-item-day level and accounting for fresh product lifecycles, promotional lift, and weather-driven demand, then turning that into order proposals a human can sanity-check on a phone in the aisle.
This is where most homegrown systems and bolted-on ERP demand modules fall down. They can forecast a distribution center reasonably well, then collapse at the shelf edge where the actual waste and lost sales live. Shell's decision reflects a common realization among operators: the last mile of forecasting is a specialist discipline. Getting it wrong on fresh means either empty shelves that send customers to a rival forecourt or markdown bins full of product that never should have been ordered.
The Buy Decision Behind a 550-Store Rollout
For a CIO, the interesting part is not that Shell wants better forecasts. Everyone wants better forecasts. The interesting part is that Shell chose to buy a purpose-built retail planning platform rather than extend an internal system or lean harder on a broad ERP suite. Janine Albrecht Webb, UK General Manager for Shell Mobility and Convenience, put it plainly: RELEX gives the company the tools to plan smarter across its UK convenience network. That is the language of a team that decided the capability was too specialized and too central to rent piecemeal.
The build-versus-buy math here is not subtle. Replicating multi-echelon, weather-aware, fresh-aware forecasting in-house is a multi-year data science and engineering commitment, and it competes for the same scarce talent every enterprise is fighting over. Buying a mature platform converts that into an integration and change-management project. The risk shifts from can we build it to can we operationalize it, which is a risk most retail operations leaders would rather own. Shell's move gives peer operators a clean reference point for that conversation.
Speed to Value Is the Real Pitch
RELEX co-founder and Group CEO Mikko Karkkainen leaned into deployment speed rather than model sophistication, saying speed matters in a network like Shell's and that RELEX can get them live and deliver value far faster than a transformation of this scale would normally take. That is a deliberate message to buyers scarred by planning programs that ran for years and delivered little. When the vendor's headline claim is time to value rather than forecast accuracy, it tells you what has actually been killing these projects.
Enterprise leaders should read that as both a promise and a test. Fast go-live is worth real money, because a planning system that ships value in quarters instead of years changes the business case entirely. It also raises the bar on your own readiness. Speed on the vendor side only pays off if your master data, store hierarchies, and promotion calendars are clean enough to feed the models. The transformations that stall rarely fail on the algorithm. They fail on the data plumbing underneath it.
Fresh, Weather, and the Margin That Hides in Waste
The capabilities Shell named are a tell about where the value is. Fresh optimization, fresh markdowns, and weather-based forecasting are all aimed at the same enemy: perishable waste and the lost margin that comes with it. In convenience retail, fresh is the category that drives repeat visits and also the one that quietly destroys profit when ordering is even slightly off. Shrinking that gap by a few points across 550 stores compounds into a number that shows up in the P and L, which is why fresh is the beachhead for almost every serious replenishment program.
Mobile replenishment is the operational glue. Order proposals generated centrally still route through store teams who can adjust before confirming, which keeps human judgment in the loop without asking staff to build orders from scratch. For operations leaders, this is the sane version of automation: the system does the heavy computation, and the person closest to the shelf keeps veto power. It is also the model most likely to survive contact with frontline reality, because staff trust a tool that assists them more than one that overrides them.
What This Means for Your Planning Roadmap
The Shell deal is a marker for anyone weighing how to modernize demand planning. The pattern is consistent across recent retail moves: operators are concluding that forecasting and replenishment are specialist capabilities worth buying, and that the differentiator is execution speed plus fresh-grade accuracy rather than a novel model. If your convenience, forecourt, or small-format estate is still planned on category averages and manual overrides, this is the competitive benchmark you are now measured against.
The decision you own is not whether AI belongs in replenishment. That argument is over. It is whether you can get your data and processes ready fast enough to make a bought platform pay off, and whether your store teams will adopt tools that change how they order. Shell has signaled that a 550-store network can commit to that path. The roadmap question for peers is sequencing: prove the fresh case in a subset of stores, get the master data clean, then scale. The technology is ready. Readiness on your side is the variable.
