Dark Stores and Dark Warehouses: The Global Story

What dark stores and dark warehouses actually are

A dark store is a retail-sized facility that is closed to the public and set up purely to pick online orders. The word “dark” has nothing to do with lighting. It simply means the shopfront is shut to customers: no tills, no aisle browsers, no impulse purchases — just racks, pickers, staging bays and riders at the door. The term was coined in the United Kingdom by Tesco, which started dedicating stores to online order fulfilment in the late 2000s because in-store picking had begun to create real friction for shoppers and staff.

A dark warehouse is a bigger, usually more automated version of the same idea. Where a dark store works at supermarket scale and handles thousands of orders a week with people doing most of the picking, a dark warehouse processes tens of thousands of orders a week using robots, conveyors and software. Ocado’s customer fulfilment centres in Britain — the first of which opened in Hatfield in the early 2000s — are the best-known example, and the technology has since been licensed to grocery groups in the United States, Japan, South Korea and Australia.

None of these facilities is a new kind of shop. The reason the vocabulary matters is that the distinction between “shop” and “warehouse” decides real-world questions: whether a property needs retail planning permission or warehouse consent, how it is taxed, what a franchise agreement says about a facility operating inside a member’s territory, and who can legally open one next to a residential street. In France, the classification question shut the entire category down (more on that in Section 4).

The five fulfilment models — and why the difference decides the economics

Between “a person with a trolley in a live shop” and “a robotic warehouse” there are several distinct models, and they carry very different costs. McKinsey’s widely cited work on online grocery economics, published in May 2022, sets out five of them with indicative capacities. These are planning numbers, not promises, but they are the best public benchmark available.

All international figures in this article are converted to South African Rand (ZAR) at approximate August 2026 rates — US$1 ≈ R16.20; €1 ≈ R18.70 — for consistency. Original amounts are shown in brackets where useful.

• In-store picking — staff pick online orders from the aisles of a normal store. Low upfront cost, high operating cost. Practical up to roughly 800–1,000 orders per week per store.
• Warehouse in store — a dedicated area inside or attached to the store where pickers assemble orders. Handles about 1,000–1,300 orders per week.
• Robotic micro-fulfilment centre — automated picking close to the customer, often inside a store’s back-of-house or a small adjacent site. Up to 4,000 orders per week, delivery in two hours or less.
• Dark store — a stand-alone site laid out for manual picking, with no walk-in trade. Around 2,400–3,000 orders per week. Suited to moderate-volume catchments.
• Highly automated warehouse — full automation across storage, picking and packing. Up to roughly 30,000 orders per week. Suited to very large demand bases.

The same research puts a number on why this matters. For a roughly R1,620 online grocery basket (US$100) picked manually in a store and delivered to the door, the retailer’s net profit margin works out at about minus R210 before any customer fees are added. Picking labour and overhead account for roughly R130 of that basket and last-mile delivery for another R130. Grocers stay profitable online by charging fees and prices that cover those costs, or by shifting fulfilment to more efficient facilities.

For an owner the practical reading is simple. If your store does a few hundred online orders a week, dedicated fulfilment is likely to cost you money, not save it. If a group is doing thousands of orders a week in one dense catchment, a dedicated node starts to earn its keep. The model choice is a volume decision, not a fashion decision.

Origins: a British invention, refined by accident

The dark store was invented in the United Kingdom, and it came out of a practical nuisance rather than a grand plan. When online grocery took off in the late 2000s, Tesco found that pickers in its stores were competing with shoppers for the same aisles, causing congestion, staff frustration and out-of-stocks. Its response was to repurpose existing stores in Croydon and Aylesford as picking-only facilities — by 2009 the company was fulfilling close to half a million orders a week out of them. The industry gave these closed-off shops the name they still carry today.

“Groceries could be delivered as quickly as a taxi could be hailed.” — Nazim Salur, Getir co-founder, describing the original promise of the dark store model.

The rise and fall pattern is worth holding on to. The dark store itself never went away — Tesco’s innovation stuck. What failed was the version that tried to run the entire grocery shop through tiny neighbourhood depots on venture capital subsidies. The formats that survived are the ones attached to real grocery operations with real supply chains.

The global spread: five markets, five different stories

United Kingdom: the original model, still the deepest market

The UK remains the market where online grocery penetration is highest among major economies. Tesco’s dark stores evolved into what the company calls dotcom centres, while Ocado’s automated warehouses became a licensable product. The model is mature enough that the argument has moved from “can online grocery work” to “which fulfilment mix works where”.

Europe: the quick-commerce boom and bust

The pandemic wave brought Gorillas, Getir, Flink and a dozen smaller operators into Berlin, London, Paris and Amsterdam with 10-minute delivery promises. The money ran out before the economics worked. Getir acquired Gorillas in December 2022 and then withdrew from Europe and the US in 2024, citing unsustainable costs. France went further: from 1 July 2023, dark stores could no longer be classified as shops, only as warehouses — a change that raised their tax and compliance burden sharply. Getir and Gorillas left France that month, with about 1,300 jobs lost. At their peak, twelve operators ran around 80 locations in Paris alone. Amsterdam, Barcelona and Paris also moved to bar dark stores from residential areas. Germany’s Flink is the notable survivor: in March 2026 it reported reaching EBITDA profitability on average baskets above roughly R840 (about €45) delivered in around 30 minutes.

United States: the retreat from the robot warehouse

America went a different route. Rather than small urban dark stores, Kroger and others invested in large automated customer fulfilment centres built with Ocado’s technology — the plan called for twenty, and eight were opened. In November 2025 Kroger said it would close three of the eight after they failed to meet financial expectations, and it expanded its partnerships with Instacart, DoorDash and Uber Eats instead, fulfilling more orders from existing stores. Reuters reported Ocado would receive over R4.1 billion (US$250 million) in compensation. Wharton professor and operations researcher Gad Allon has argued the failure was structural: automated picking cuts warehouse costs, but last-mile delivery remains the dominant cost, and sprawling US geographies punish centralised warehouses with long, unprofitable routes. The American lesson is that automation solves the picking problem, not the delivery problem.

China: the 30-minute standard

Chinese players ran ahead of the rest of the world on speed. Dingdong Maicai, founded in Shanghai in 2017, built a network of neighbourhood “front-line fulfilment stations” — dark stores in all but name — promising 30-minute delivery, and reported its first full year of net profit in early 2025. Meituan and Alibaba’s Hema operate comparable models at scale. The failures are instructive too: Miss Fresh, an early pioneer, shut down after burning through billions, and community group-buying formats shrank sharply. Speed is standard in China; profitability was the hard part.

India: the biggest and fastest dark-store build-out

India has become the most aggressive dark-store market in the world. The quick-commerce sector reached roughly R123 billion (₹64,000 crore, about US$7.6 billion) in gross order value in FY2025, more than double the previous year, and now accounts for around a third of online FMCG purchases in some urban households. Blinkit (owned by Zomato), Zepto and Swiggy Instamart run networks of around 600 to 1,100 dark stores each, and the buildings are growing: Zepto is opening stores of up to 8,000 square feet to hold larger items such as appliances. The model has also drawn regulatory attention, with distributors filing a predatory-pricing complaint with India’s competition commission in 2024. Analysts expect the market to keep growing to around R210 billion (about US$13 billion) by 2029, driven by advertising revenue on top of thin fulfilment margins.

What the model does well — and where it fails

Set the marketing aside and the evidence on dark stores is fairly consistent across markets. They are good at a specific set of things and bad at almost everything else.

Where the evidence is in its favour

Dedicated facilities take pickers out of the aisles, which was Tesco’s original reason for building them. That removes congestion, protects shelf availability for walk-in customers, and lets staff work at higher pick speeds with fewer interruptions. The second benefit is throughput per square metre: McKinsey’s benchmarks put a manual dark store at 2,400–3,000 orders a week, roughly three times an in-store operation. The third is availability for online customers. Woolworths said its first dark store in Cape Town’s city bowl significantly improved product availability for Dash orders, and the Wynberg facility is explicitly aimed at improving access to delivery slots and order accuracy. A dedicated stock pool, planned against online demand only, is easier to keep in stock than a store serving two masters.

There is also a control argument. A closed facility is easier to run at a consistent standard — temperature control, substitution rules, age verification on alcohol, CCTV — than a live store where the same staff are juggling customers. That matters more as baskets move into liquor and pharmaceuticals, as they have in South Africa’s on-demand market.

Where the evidence is against it

The baseline economics are unforgiving. McKinsey’s analysis puts a manually picked and delivered R1,620 basket (US$100) at roughly minus R210 of net margin before fees, and while dark stores improve on that, they only do so at volume. Below a few hundred orders a week, the fixed costs of a dedicated facility — rent, fit-out, a second inventory pool, dedicated staff — outweigh the picking savings. The model is density-dependent by construction.

Kroger’s experience shows the risks at the other end of the scale. Its automated warehouses, built on cheaper land outside cities, cut picking costs but lengthened delivery routes, and three of the eight failed to meet financial expectations. Automation solved the warehouse problem and left the delivery problem untouched. The same pattern ended the European quick-commerce experiment: in low-density neighbourhoods, tiny dark stores could never cover their rent, staffing and rider costs at the prices consumers would pay.

Regulation and property add a second layer of risk. France’s reclassification of dark stores as warehouses, and the residential-area restrictions in Amsterdam, Barcelona and Paris, show that the model depends on where planning law lets it exist. A facility that operates in a planning grey zone is a facility with an expiry date.

The seven pressure points every owner should understand

Anyone evaluating dark stores — as an operator, a franchisee or a competitor — ends up working through the same seven issues. This section keeps each one short and practical.

6.1 Operations

A dark store is a picking machine, and its layout, stock file and staging area determine its throughput. Replenishment is the silent problem: every dedicated facility is a second inventory pool, supplied separately from stores, and it needs its own forecast, its own deliveries and its own waste. Peak periods — Friday evenings, weekends, holidays — determine whether the operation works, because that is when capacity runs out.

6.2 Finance

Capital goes in before revenue shows up: lease, fit-out, racking, chillers, software, working capital for a full second stockholding. The per-order economics decide whether the facility ever pays for itself, and they are dominated by the two costs McKinsey identifies — picking labour and last-mile delivery. Groups also carry the write-off risk Kroger demonstrated: facilities that cannot reach their order volumes become expensive mistakes.

6.3 Staffing

Dark stores need pickers, dispatchers, quality checkers and managers, and the work is physical and repetitive. Rider and driver fleets raise a separate set of questions about employment classification, safety and liability — Sixty60’s roughly 10,000 Pingo drivers in South Africa are contracted through the group’s own subsidiary, which Shoprite took into full ownership in October 2024 to tighten standards. In franchise settings, who employs the pickers and who carries the labour-law exposure is a contractual question with real money behind it.

6.4 Technology

The app, the warehouse management system, the routing engine and the substitution logic are the invisible backbone. South Africa’s example is instructive: Zulzi built the original Sixty60 platform and now runs its own seven dark stores in Johannesburg and Pretoria, showing how much of the value in this model sits in software rather than shelves. Forecasting tools that can predict demand by hour, and substitution rules that customers accept, are what separate a smooth operation from a complaints queue.

6.5 Delivery

Last mile is the cost nobody automates away. Fuel, rider pay, average distance and orders per hour determine it, and in South Africa load-shedding adds a layer of uncertainty that northern-hemisphere operators never think about. Sixty60’s scale — the platform now accounts for more than 10% of Shoprite’s South African supermarket sales — shows what a dense network can achieve, and how far a two-hour delivery promise can feel in comparison.

6.6 Property

Location is the whole game: a dark store must sit close to demand, which means urban and suburban retail or industrial space at urban prices. Planning classification decides what can open where, as France demonstrated, and residential opposition has forced closures in several European cities. Lease terms matter too — a facility fitted out for picking has little alternative use if the venture fails.

6.7 Customer service

Substitutions, missing items and cold-chain failures are where online grocery loses customers, and the pressure rises as delivery promises tighten. Service standards — how substitutions are offered, how quickly complaints are handled, whether the cold chain survives the doorstep — are measurable and should be managed as tightly as store standards. The flip side is data: the customer record from an app order belongs to whoever runs the app, and that determines who can market to that customer tomorrow. That single question matters as much as any technology in this model.

Franchisees: what the evidence shows — and what it does not

This section deserves care, because the honest answer is that the evidence on how dark stores affect supermarket franchisees is thinner than the commentary suggests. Here is what is actually documented, followed by what is not.

What is documented

The clearest global case of online grocery run through franchise stores is SPAR in South Africa. SPAR2U is fulfilled from participating SPAR and TOPS stores operated by independent retailers. The group reported the service live in 581 sites by March 2025 — up from 420 a year earlier — with delivery volumes up 174% year on year, and by September 2025 it was active at 636 sites with total order volumes up 136% for the financial year. An Uber Eats integration, launched in early 2025, was live at more than 300 sites, with products listed at retail price plus Uber’s take rate of around 25–30%.

SPAR’s own framing has been that the platform takes “all of the onerous effort out of our retailers’ hands”. Retail analyst The Finance Ghost, by contrast, called SPAR2U “a complete non-event”, pointing to the difficulty of getting independent retailers to participate and of accessing data across the network. Both positions are on the record.

The group’s financial pressure is also on the record, but the causes are documented as specific. In H1 FY2026, SPAR’s operating margin fell to 1.1% (1.3% before exceptional items) from 2.1% a year earlier, with headline earnings per share down 55.5%. The company attributed the decline to underperformance at its KwaZulu-Natal distribution centre, an ineffective Black Friday campaign and balance-sheet clean-ups — not to dark stores. New group CEO Reeza Isaacs, who took over in March 2026, has said the problems are execution problems rather than market problems, and has flagged “repositioning SPAR2U” as one of five recovery priorities. It would be wrong to claim dark stores caused SPAR’s margin decline; the company’s own account points elsewhere, and the retail environment has been tough for everyone.

Where the structural analysis points

AnalysisThere is a structural difference between how corporate and franchise networks absorb online grocery, and it is visible in the South African contrast. Checkers and Woolworths pick most of their online orders in company-owned stores. If an order moves to a dark store, the same balance sheet gains the picking efficiency and the same balance sheet loses the in-store turnover — the group reallocates rent, labour and stock across its network. In a franchise network, the store owner supplies the stock, floor space, fridges, picking labour and shrink, while the franchisor typically controls the app, pricing, promotions and customer data. When online volume is proven in a franchisee’s catchment and then moves to a group-run facility, the franchisee’s fixed costs do not move with the turnover. That is a real risk embedded in the structure of the model, not a statement that it has already happened.

SPAR’s own recovery plan is, tellingly, built around retailer profitability as its primary metric — benchmarking, staff-scheduling tools, rental-negotiation support and a repositioned digital platform. That is the strongest available signal from a franchise-based group about where the pressure sits.

Outside grocery supermarkets, the franchise version of quick commerce exists too: The Grocer reported in 2024 that Getir operated through franchised One Stop stores in parts of the UK, alongside its own depots. The model is not unique to SPAR.

The contractual questions worth putting in writing

AnalysisMost supermarket franchise agreements were written before on-demand grocery existed. The questions that matter, in no particular order, are:

• Whether online sales delivered into a franchisee’s territory are recognised as the franchisee’s turnover, the franchisor’s, or a hybrid.
• What the picking fee or fulfilment recharge actually covers — labour, packaging, space, utilities, shrink — and how often it is reviewed.
• Who owns the customer record and transaction data from orders picked in the franchisee’s store.
• Whether the franchisor may open a dark store or appoint a third-party fulfilment partner inside the franchisee’s territory, and with what notice or compensation.
• How online promotions are funded when the app discounts a line the store was not planning to discount.

These are due-diligence questions, not predictions. Franchisees who can answer them on paper will be in a better position than those who discover the answers after a facility opens nearby.

South Africa: the regional test case

South Africa is one of the clearest live examples of the global pattern, because the country skipped much of the European detour and went straight from in-store picking to dedicated nodes in dense catchments.

The market context

ConfirmedOnline retail in South Africa reached R96 billion in 2024, up 35%, and the World Wide Worx, Mastercard, Peach Payments and Ask Afrika report projects it will pass R130 billion in 2025 — nearly 10% of total retail — and exceed R150 billion, about 12%, by 2027. On-demand grocery is a large part of that growth.

Checkers Sixty60

ConfirmedSixty60 launched in November 2019, just before the pandemic, and has grown into South Africa’s dominant on-demand grocery platform. It turned over R11.9 billion in the six months to December 2025, up 34.6%, and R18.9 billion for the 52 weeks to June 2025, up 47.7%. By late 2025 it was fulfilling from roughly 875 stores using about 10,000 Pingo drivers, had passed 100 million cumulative orders in March 2025, and accounted for more than 10% of Shoprite’s South African supermarket sales. The group took full ownership of Pingo in October 2024, bringing the last-mile layer in-house.

ConfirmedCheckers’ first dedicated dark store was planned for Bree Street in Cape Town back in 2022; the confirmed openings followed in Gardens and Maitland in November 2024, and by July 2026 the group had six dark stores across the Western Cape, Gauteng and KwaZulu-Natal, with Midstream Estate near Centurion among the sites reported.

Woolworths Dash

ConfirmedWoolworths’ on-demand Dash service launched in December 2020. In FY2025, Dash turnover grew 41.6%, overall online sales reached R7.78 billion (up 8.9%), and online food sales of R3.33 billion (up 32.9%) represented 6.6% of total food sales. Woolworths opened its first dark store in Cape Town’s CBD in August 2024 and a second in Wynberg in June 2026, and has said more are planned. The group’s supply chain chief, Bradley Nitsckie, has said the facilities are aimed at improving speed, availability and delivery-slot access.

SPAR2U and the franchise dimension

ConfirmedSPAR2U’s numbers are covered in Section 7: 581 sites by March 2025, 636 by September 2025, volumes up 174% in H1 FY2025 and 136% for FY2025, with an Uber Eats channel at 300+ sites. The platform is fulfilled from franchisee stores, which is what makes it the country’s clearest test of the franchise model in online grocery.

The smaller players

ConfirmedZulzi, the company that built the original Sixty60 platform, operates seven of its own dark stores across Johannesburg North, Pretoria and Centurion, carrying grocery, liquor and pharmaceutical lines. Pick n Pay’s asap! service runs through more than 620 stores with over 2,500 drivers, and the group said its online business remained profitable in FY2026 even as growth slowed.

Opportunities and risks: franchisees, independents and groups

The balanced reading, based on the evidence gathered above, looks like this. Dark stores are neither a threat to every supermarket nor a guaranteed advantage for the operators who build them. The outcomes depend on who is building, in what catchment, and under what contract.

The evidence from India shows that the biggest growth in this model is real — but it also shows the two risks that travel with it. The first is that thin fulfilment margins push operators toward advertising revenue, which changes who the platform ultimately serves. The second is regulatory and political pushback when dark stores are seen to be squeezing small traders, which is exactly what the AICPDF complaint in India and the French reclassification both represent. None of this makes dark stores a passing fad. It does mean the model is being shaped by regulation, consumer behaviour and unit economics at least as much as by technology.

A practical checklist for owners evaluating a dark-store strategy

Whether you are a franchisee deciding how to respond to a group platform, an independent deciding whether to invest, or a group executive approving a network, the same eight steps apply. They are deliberately short on theory and long on arithmetic.

None of this is a prediction. The global record shows dark stores and dark warehouses are here to stay in the markets and catchments where the economics work, and that the model keeps getting reshaped by regulation, capital discipline and consumer behaviour. For a supermarket owner, franchisee or independent, the useful question is not whether dark stores are good or bad. It is whether your catchment, your contract and your cost base put you on the right side of the arithmetic.