A supermarket can be certain it owns twelve tubs of yogurt and still have none a shopper can buy. The stock may have been stolen, broken, scanned incorrectly, left in a cage out back or sold without the inventory file getting the memo. The database remains serene. The shelf is bare. In retail argot, those ghost tubs are phantom inventory. In human terms, somebody leaves without breakfast.
Retail Insight has spent two decades in that discrepancy. Founded in Britain in 2005, the company takes the data grocers and consumer-goods companies already generate - sales, stock, waste, promotions, online orders - and converts it into a short, ranked set of interventions. Restock this line. Check that count. Mark down those sandwiches now. Fix the promotion in aisle nine. The software runs in the cloud, but its moment of truth is pleasingly physical: a store colleague looks at a shelf and does something.
That is a narrower proposition than “transform retail with AI,” and more useful. Retailers already possess warehouses full of reports. They have fewer people in stores, tight margins and a daily queue of jobs that could swallow every shift. Retail Insight sells the translation between abundant information and scarce attention.
Company-reported current metrics
The first thing to fail was the inventory file
Asda offers the cleanest view of the machinery. Retail Insight had worked with the British grocer on availability since 2009. Yet store teams still encountered discrepancies while replenishing shelves. An item could show positive stock in the system even when none was sellable. The causes were prosaic and numerous: theft, short deliveries, poor process compliance and checkout errors.
The companies trialed InventoryInsight in February and March 2023. The model used sales and inventory data to identify likely phantoms, ranked the high-impact cases and sent alerts to store colleagues for shelf-edge validation. A person checked the claim; the inventory record was then corrected. After the trial, Asda chose a full rollout beginning that July. This is the “what changed their mind” moment: not a grand AI vision, but alert accuracy, recovered availability and a workflow that store teams could actually follow.
In under twelve months, the case study says, Asda removed 1.6 million phantom units. Each worked alert removed 19 units on average, and the project returned 63 times its cost. That last figure is unusually muscular for enterprise software, so it deserves its proper label: a result reported by Retail Insight and its customer, not a universal promise. The financial price of the deployment was not published.
“The decision to rollout InventoryInsight for our stores was a no-brainer given the strong results we saw in the trial period.”Neil Fairclough, Senior Director of Retail Transformation, Asda
A product family named after the mess
The product names are wonderfully literal. AvailabilityInsight finds likely out-of-stocks and tells staff where intervention will recover the most sales. WasteInsight calculates markdowns for expiring products and now also handles stock exits, donation opportunities and food-safety warnings. OnlineInsight spots digital availability risk, recommends less absurd substitutions and diagnoses the journey from order to pick. InventoryInsight hunts the ghosts. PromoInsight checks whether offers made it from the marketing plan to the shelf. StoreInsight prioritizes the work of running an estate.
AvailabilityInsight
Which missing item is worth fixing first?
WasteInsight
When should an expiring item be reduced, donated or removed?
OnlineInsight
Which order is at risk, and what substitute makes sense?
InventoryInsight
Which stock record is confidently wrong?
PromoInsight
Which planned offer is failing in the store?
NRMInsight
Can brand and retailer agree on one trading P&L?
For consumer-goods companies, ExecutionInsight and DART turn retailer data into priorities for commercial and field teams. NRMInsight models trade investments, promotions and profit-and-loss lines so a brand and retailer can build a joint plan from the same numbers. The company says more than 300 CPGs use its profit model across more than ten markets, with a typical tenfold customer return.
Its moat is a memory of stores
Retail Insight competes in a crowded aisle. RELEX Solutions, Blue Yonder and SymphonyAI sell broad retail planning and supply-chain suites. Mi9 Retail covers merchandising and store operations. Whywaste and Wasteless attack portions of expiry management and markdowns. Retailers can also build models on general cloud platforms, hire consultants or keep extending the reports they already own.
Retail Insight's answer is not that its mathematics came from another planet. It is that the models have been trained for retail, sit above existing technology and arrive with people who know how stores behave. The company claims more than 500 years of collective retail and CPG experience. Its current footprint gives those models another advantage: the company says it processes $1.4 trillion of grocery revenue and holds 15 percent of global grocery data on its platform.
The more important difference is product philosophy. A planning suite helps decide what should happen across a network. Retail Insight is strongest at the ragged edge where a plan meets Tuesday afternoon: one product, one store, one colleague, one action. It calls the destination “shelf actualization,” a slightly lofty name for the practical state in which products are available, inventory is credible, promotions work and expiring food finds a buyer.
The customer list is the argument
Public customers include Asda, Kroger, Sprouts Farmers Market, Marks & Spencer and several Co-op organizations. Its CPG footprint includes names such as Mondelez, Colgate-Palmolive, Unilever and Duracell. In 2021, Kroger piloted AvailabilityInsight in two divisions, then expanded the collaboration toward stores across the United States. Sprouts selected the same product for 380 stores. Asda commissioned five handheld applications spanning availability alerts, stock picking, binning, gap scanning and bin audits.
Waste supplies the sharper emotional hook. A markdown is both a margin decision and a race against rot. Retail Insight says its platform kept 856 million meals from landfill last year and reduced 1.4 million tonnes of carbon dioxide. In a 2025 award submission, Southern Co-op reported diverting more than 3,000 tonnes of food from landfill in 2024, saving 3.08 million meals and increasing reduced-price sell-through by 17 percent. The commercial and environmental incentives happen to share a barcode.
“Our partnership with Retail Insight supports us in achieving our goal to halve food waste by 2030.”Nick Meagher, Head of Analytics & Insight, Co-op
Software, specialists and a patient sales cycle
This is enterprise B2B software rather than a swipe-the-card subscription. Retail Insight does not publish prices. It sells modules, implementation and domain expertise to large retailers and manufacturers, usually through a scoped business problem. OnlineInsight, for example, advertises a pilot in five to ten stores, deployment in six to eight weeks and measurable results within seven days. The point is to make the buying decision reversible before asking a sprawling estate to change its habits.
The company itself has followed patient capital. Paul Boyle, an engineer who previously worked at P&G and Heinz, served as COO from the 2005 founding and became CEO in 2011. Navis Capital Partners backed the business before Ventiga Capital Partners bought it in July 2019. In June 2024, ThinCats provided an undisclosed debt package to refinance existing borrowing and support working capital. The company later added product chief Alex Considine Tong and technology chief Fredrik Palm. In 2025, it was named Data and Analytics Company of the Year at the Retail Systems Awards.
ERP, point-of-sale, planning and replenishment systems create the operational record.
Retail-trained models detect anomalies, estimate value and rank the next action.
Broad suites such as RELEX, Blue Yonder and SymphonyAI overlap in planning and execution.
Store colleagues verify the shelf, change a label, correct a count or replenish a line.
The part worth copying is the loop
A retailer cannot copy Retail Insight's data estate overnight. It can copy the operating logic. Begin with a loss that staff recognize. Define it at item-store-day level. Rank interventions by value rather than sending every possible alert. Put the recommendation inside the workflow people already use. Require a human verification where the data is uncertain. Record what happened. Feed the result back into the model. Expand only after the pilot survives contact with a busy store.
- Choose one expensive decision, not a vague transformation.
- Measure action taken and value recovered, not dashboard visits.
- Keep the alert queue short enough to earn attention.
- Use pilots to test process compliance as well as model accuracy.
- Scale when colleagues trust the recommendation and managers can see the result.
The conditions matter. These systems need timely foundational data, a repeated decision and somebody empowered to act. A brilliant alert that arrives after the delivery order closes is a historical curiosity. A correct markdown recommendation without labels, labor or authority is merely arithmetic. And a torrent of low-value alerts will train staff to ignore the machine. Retail Insight's best case studies work because the software, the store process and the measurement were designed as one loop.
That may be the company's least flashy and most defensible insight. The grocery store is not short of intelligence. It is short of clean handoffs between a system that notices, a person who trusts it and a shelf that needs fixing before dinner.