A forecast is a wonderfully confident object. It turns next month into a neat column of numbers, then waits for a port strike, a heat wave, a celebrity TikTok, or one missing microchip to make it look silly. ToolsGroup has built a 33-year business around a less comforting idea: the future is not a number. It is a range of possible outcomes, each carrying a different cost if a company stocks too much, too little, or in the wrong warehouse.
That sounds academic until the missing item is an Aston Martin wing mirror. The carmaker’s aftermarket operation had 80,000 spare-parts SKUs, international customers with little patience, and a board unwilling to solve availability by piling up expensive safety stock. ToolsGroup’s software grouped demand into eight newly detected patterns, tuned safety stock daily, and generated replenishment plans. Within two months, Aston Martin reported an 18 percent reduction in safety-stock value on those clustered items while first-time availability reached 97.1 percent.
What failed first? The single-number forecast.
Joseph Shamir and Eugenio Cornacchia founded ToolsGroup in Boston in 1993 after conceiving the company on a train ride between Milan and Genoa. Their premise was that demand and supply behave probabilistically. Traditional planning systems were excellent at transactions - recording orders, moving goods, closing the books - but weaker at deciding how much uncertainty to buffer at each location. ToolsGroup built service-driven planning around that gap.
Instead of asking only, “How many will we sell?”, the system asks a more useful sequence: How wide is the range? What service level has the business promised? Which SKU-location combinations deserve scarce working capital? How does a stocking decision at one node alter risk elsewhere in the network? This is multi-echelon inventory optimization in plain clothes: do not optimize each warehouse as an island, and do not treat every item like the average item.
“The great thing about the eight categories is that people can see them.”Nick Wilson, former senior inventory planner, Aston Martin Parts Operations
The visibility mattered. A planner asked to trust an opaque recommendation will keep a spreadsheet nearby like a security blanket. At Aston Martin, recognizable demand categories taught the purchasing team something about its own catalogue. The machine’s first persuasive act was not automation. It was legibility.
Other customer accounts repeat the basic trade. Boise set out to cut inventory 10 percent while improving delivery, then reported more than twice that inventory reduction and a three-point service improvement in year one. Bellota said worldwide availability rose from 93 to 96 percent in four months while inventory fell 14 percent. These are company-published case studies, not controlled experiments; they still show the metric pair ToolsGroup wants buyers to watch. Inventory alone rewards empty shelves. Service alone rewards hoarding. The useful number lives between them.
The product became a portfolio, then a platform
For years, the flagship name was Service Optimizer 99+, or SO99+. ToolsGroup widened the aperture after Accel-KKR first invested growth equity in 2018. It acquired the demand-management business formerly known as JustEnough in 2021, adding retail assortment, allocation, and merchandise planning. Onera followed in 2022 with real-time inventory availability and fulfillment. Evo arrived in 2023 with price, promotion, markdown, and inventory optimization.
The acquisitions answered an awkward problem in enterprise planning: a clever inventory plan can be spoiled by a pricing move, a store allocation, or a fulfillment decision made somewhere else. In 2024, Decision Hub added a virtual situation room where cross-functional teams could compare scenarios. In May 2026, ToolsGroup placed the accumulated engines under one name, Decion. Existing SO99+, JustEnough, and Evo capabilities carried forward inside a unified platform.
How the decision surface expanded
Conceptual map of publicly announced product scope, not a revenue or market-share chart.
Decion covers demand and supply planning, price and promotion optimization, strategic planning, retail planning, and visibility. Its “Decy” assistant can surface risks, explain recommendations, and support scenarios. The more consequential claim is controlled execution: agents can automate approved actions inside business guardrails. The planner moves from touching every replenishment proposal to defining tolerances, investigating exceptions, and choosing which decisions deserve a human.
That places ToolsGroup in a crowded market beside SAP, Oracle, Blue Yonder, Kinaxis, o9, Logility, Infor, OMP, and a lively pack of specialists. The company is narrower than the biggest enterprise suites and broader than a point forecasting tool. Its differentiator is not merely “AI,” a label now applied to nearly every planning screen. It is the combination of probabilistic modeling, service-level economics, multi-objective optimization, and workflows intended to move a recommendation toward action.
Who buys it - and what they are buying
ToolsGroup says more than 400 companies in over 45 countries use its software. The sweet spot is operational complexity: manufacturers with component constraints, distributors with vast long-tail catalogues, retailers balancing stores and channels, food businesses worried about freshness, and aftermarket teams where a rarely ordered part can carry an outsized service promise. The customer list includes names such as Toyota, Thule, Harley-Davidson, Absolut, Aston Martin, Boise, and Bellota.
This is enterprise SaaS, sold by subscription with implementation, professional services, support, and partner delivery around it. ToolsGroup had become predominantly SaaS by 2018 and was profitable at the time of Accel-KKR’s first investment. The economics are straightforward even when the implementation is not: free working capital trapped in inventory, protect availability, reduce expedites and markdowns, and let planners spend less time nursing routine orders.
Microsoft is a notable technology partner; ToolsGroup moved cloud delivery to Azure and has sold through Microsoft channels. Its wider network counts more than 40 consulting, technology, and reseller partners. That ecosystem matters because planning software sits between ERP records, commercial assumptions, supplier constraints, warehouse reality, and human incentives. Installation is the easy noun. Adoption is the verb.
The part worth stealing
A large electronics distributor offered perhaps the cleanest playbook. Other vendors would not run a proof of concept on the requested scale. ToolsGroup tested 700,000 SKUs with varied service levels. The solution went live in February 2021 and reached all planners by April. The customer later reported canceling $343 million in inventory orders, pushing out another $300 million, and reducing proposals by $20 million while stabilizing safety stock. The important sales move was agreeing to be judged on the customer’s ugly reality, not a polished sample.
Use the long tail, intermittent demand, and awkward locations in the proof of concept. Easy SKUs prove very little.
Measure service and inventory together. Add planner workload, expedites, margin, or waste where they matter.
Show demand behaviors, trade-offs, and recommendation logic so operators can challenge the model intelligently.
Begin with recommendations and bounded exceptions. Expand autonomy only after accuracy and ownership are clear.
This is also what changed minds in earlier deployments. ADI began with a pilot before moving to a full implementation during a logistics-network redesign. Aston Martin had used ToolsGroup for a decade before adding the new machine-learning engine. Trust accumulated through observed decisions. The copyable idea is almost quaint: earn the right to automate.
When the playbook breaks
Probabilistic optimization cannot rescue unreliable item and location data, undefined service policies, or a team that will override every unfamiliar answer. It can also be excessive for a simple network with stable demand and low inventory stakes. If pricing, operations, and finance refuse to agree on goals, a multi-objective engine merely receives a more sophisticated version of the argument.
The bet behind “self-steering”
Sean Elliott took over as chief executive in February 2025 after 17 years at Körber Supply Chain Software. His timing is convenient and difficult. Generative AI made assistants easy to demonstrate; supply chains make autonomous actions hard to forgive. Ordering the wrong quantity is not a hallucinated paragraph. It is a truck, an invoice, a full rack, and perhaps a markdown six months later.
Decion’s wager is that agentic software becomes useful when joined to the older, less fashionable machinery: probability distributions, constraints, optimization, service targets, and approval rules. The assistant is the conversational surface. The durable product is the decision system underneath it.
“Decion represents a fundamental shift for the industry - from predicting what might happen to continuously steering decisions towards the right business outcomes.”Sean Elliott, chief executive, ToolsGroup
There is a pleasing loop to the story. ToolsGroup began by refusing the false certainty of a single forecast. It now wants customers to accept a different kind of uncertainty: software that does not wait politely for the monthly planning meeting. Whether Decion earns that freedom will depend less on how fluently an assistant talks than on whether its next recommendation is visible, bounded, and economically right. In a warehouse, charisma has never moved a pallet.