Retail fulfilment models: store, dark store, warehouse or 3PL

How the comparison was made

The unit of comparison is an order from stock reservation to customer handover. Documents were included when a platform owner, retailer or standards body described allocation, inventory, fulfilment or traceability. The source ledger records URLs, passages, access limits and the claim each supports. Product documentation establishes specific functionality; a retailer report establishes a reported operating example. Neither supplies a controlled comparison of costs. There are no interviews, timed picking observations or transaction datasets. The worksheet separates documented features from hypotheses requiring a local pilot.

Picking from the trading floor

Store picking uses an existing assortment and a location already serving customers. Its operating question is how online work shares shelves, replenishment and staff with store trade. Measure the complete picking cycle, including searching for missing items, approving substitutions and waiting for packing space. An apparently short picking task can create a queue at dispatch. Use separate observations for quiet and busy trading periods. The analytical advantage is access to existing facilities; the unresolved issue is usable capacity while maintaining the store service expected by walk-in customers.

A dedicated store zone or dark store

A dedicated zone changes the picking layout and allocation of space; a dark store removes public shopping from the facility. Both require deliberate replenishment and an assortment sized for the orders they will serve. Tesco’s 2026 annual report describes a store converted to include an urban fulfilment centre, demonstrating that a mixed arrangement exists. It does not establish the economics of a new site. Test whether fewer interruptions compensate for replenishment work, additional stock and the premises or selling space assigned to online orders.

A warehouse serving online orders

A warehouse allows the retailer to design receiving, storage, picking and dispatch around fulfilment. Its suitability depends on assortment, order shape and transport geography. A facility configured for replenishing stores may need different handling for individual customer baskets. Compare travel distance, packing capacity, cut-off times and the last-mile journey. Shopify’s order-routing documentation illustrates the trade-off between keeping an order together and selecting fulfilment locations. Routing rules are a software mechanism; the retailer must still establish whether each location can perform the work promised.

Third-party fulfilment

A 3PL arrangement transfers specified warehouse operations to a partner. Amazon describes MCF as storage, picking, packing and shipping for orders from other sales channels, with an inventory pool that can also serve FBA. This confirms an available service pattern, with eligibility dependent on the market and goods. For a procurement decision, request the actual rate card, receiving rules, peak commitments, returns handling and exit terms. Compare the full invoice with internal costs over the same order mix; keep carrier charges, exceptional handling and minimum commitments visible.

Inventory and the customer promise

Shopify distinguishes stock physically present from quantities available to sell, committed or unavailable. Its local-delivery rules require the full eligible order to be available at one location. These are useful examples of why a network stock total cannot automatically support a basket promise. GS1’s EPCIS guideline describes event data about objects, time, location and business context. Applying that principle, record reservation, picking completion and carrier handover with consistent identifiers. A shared event vocabulary helps investigate discrepancies; physical counts validate the recorded availability.

Evidence to collect before choosing

Use the downloadable matrix to compare the same baskets, area and service window across options. Record labour minutes, space, stock, transport and exception costs separately. Define the denominator as fulfilled orders and retain the costs of cancellations in the period. Test normal and peak demand, then identify the next capacity step and its cost. Reject an option for the proposed promise when a necessary condition remains unverified. The result should be a conditional choice with an owner, evidence gaps and a pilot acceptance threshold for the selected network.

Working files

Sources

Apply this to your business

Bring your business question and current figures. We will define the work, the data required and the decisions your team needs to make.

Continue reading