Tokmanni's replenishment rollout: useful automation without proven autonomous buying

External-source research synthesis, not an OfferSignal client result. Based on the retained October 2026 research and source-specific dates.

A retailer expanding across countries needs more than a sales forecast. It must decide what each location should receive, when to replenish it and how much stock to hold without creating avoidable stockouts or excess inventory. Tokmanni Group’s announced forecasting and replenishment programme is useful because it puts machine assistance inside that planning loop—but does not show a machine choosing suppliers or paying them.

The announced Nordic rollout

On 16 October 2024, Tokmanni and RELEX announced that Tokmanni would replace a legacy supply-chain system with RELEX Forecasting & Replenishment across its Nordic distribution centres and stores, beginning with Tokmanni stores in Finland. The stated estate was three distribution centres and more than 370 stores across Finland, Sweden and Denmark. Objectives were greater replenishment automation and efficiency, better product availability, fewer stockouts and optimised stock levels. 1

These were forward-looking objectives, not performance results. Finnish logistics trade coverage dated 6 November 2024 repeated the Finland-first plan but added no go-live confirmation or field outcome. 2 Through 23 October 2026, the retained public record does not confirm the technical go-live date, current coverage or realised availability, inventory, labour, forecast-accuracy or return-on-investment result.

Tokmanni’s FY2025 report records 392 stores and group inventories of EUR 424.8 million at year-end. Those are operating-scale figures, not before-and-after RELEX baselines or effects attributable to the rollout. The FY2024 and FY2025 board reports contain no named RELEX result, rollout stage, automation rate, stockout measure or purchase-order authority. That is a bounded public-disclosure finding, not proof that the software was not used. 3 4

What the planning layer can do

RELEX’s public product reference describes machine-learning and data-mining forecasts, usually at product-location-day level, using demand patterns, promotions, price, range, space, holidays, weather and local events. Order proposals can combine forecast demand with on-hand and downstream inventory, open orders, estimated delivery dates, lead times, order calendars, batches, safety or presentation stock, capacity and supplier restrictions. 5

The distinctions matter. A forecast estimates expected customer purchases; a demand projection indicates what planners may need to order; an order proposal is not yet a purchase order. Generic RELEX documentation describes business rules, exception handling and user review or acceptance. Proposals are typically checked and approved before a customer back end converts them into purchase orders, although configurations can differ. This is vendor-wide functionality, not evidence of Tokmanni’s configuration. The case does not disclose whether the live process only creates distribution-centre-to-store replenishment, produces supplier-facing orders, auto-accepts thresholds or requires human approval.

A separate data-plumbing precedent

A Maersk case study describes a separate Tokmanni supply-chain implementation from 2020 that centralised purchase-order and shipment information, exposed vendor and carrier status, tracked progress and estimated arrival times, generated daily milestone reports and interfaced with Tokmanni IT. Maersk reports 97% of shipments loaded as booked in 2021, purchase orders doubling within a year and container utilisation improving by more than 50% after consolidation. A Tokmanni forwarding manager described shipment volume doubling in two years without increased full-time headcount. 6

These are vendor-published transport and visibility claims, not RELEX outcomes, and no assigned source establishes a Maersk-to-RELEX connection. They nevertheless show why product-location identity, inventory state, open orders, delivery estimates, lead times, pack and batch constraints, capacity and supplier restrictions are operational inputs—not optional catalogue enrichment.

An illustrative controlled application

For a retailer or distributor considering similar automation, start with one category or comparable group of stores. This is an illustrative application, not a reported Tokmanni deployment. Preserve authoritative stock, open-order and delivery data, define who owns each business rule, and keep order proposals separate from issued purchase orders. Human approval should remain explicit for exceptional value, promotions, new or discontinued items, projected stockouts, scarce capacity and late supply unless a documented authority model says otherwise.

Test a pre-specified baseline against a phased comparison. Measure on-shelf availability and the lost-sales definition, stockout or fill rate, inventory value and days, waste, manual override and exception rates, proposal-to-purchase-order acceptance, and service versus working-capital trade-offs. Forecast accuracy alone would not establish commercial improvement.

Tokmanni is therefore adjacent to machine buying rather than a demonstrated external procurement agent. The evidence supports automated prediction, optimisation and rules-based replenishment proposals. It does not support claims of a large-language-model buyer, supplier discovery, negotiation, request-for-quotation issuance, contractual award, payment or autonomous purchasing. The next useful disclosure would identify deployed scope, authoritative feeds, approval boundaries and measured Finland-first outcomes.

Sources and next reading

Explore OfferSignal · Focused Business & AI Workshop · All research

References