Flexible Fulfillment Strategies for 2026: BOPIS, Click-&-Collect, Dark Stores & Micro-Fulfillment Centers

That is not a joke about corporate retreat. It is the single most useful thing that happened in fulfillment last year, because it settled an argument the industry had been having since 2020.

The question was never whether robots beat shelves, or whether delivery beats pickup. The question is how many orders you can win inside a given radius. Every fulfillment model on the 2026 menu is really just a different bet on that one number.

Four models, one variable

The acronyms multiply faster than the ideas, so it helps to strip them down. Every fulfillment option answers two questions: where does the inventory sit, and who pays for the last mile.

BOPIS and click-and-collect put the inventory in a store you already own and hand the last mile to the customer, who does it for free and often buys something else while they are there. Dark stores put dedicated stock in a small local unit closed to the public, and you pay for the last mile in exchange for speed. Micro-fulfillment centers bolt automation into or beside a store so the pick itself gets faster, feeding either pickup or delivery.

Three different answers, three different cost structures. And one variable sitting underneath all of them: order density.

Everything else is downstream.

Why pickup quietly won

Among US grocery shoppers, pickup has now overtaken same-day delivery, 31% to 29%, and more than 90% of consumers shop groceries both online and in store. Click-and-collect is not the compromise choice anymore. For planned, routine baskets, it is the preferred one.

The consumer logic is not mysterious. There is no delivery fee. You choose your own timing instead of waiting inside a window. And you can grab the thing you forgot on the way out.

In a market where shoppers are scrutinising every charge, free fulfillment stopped being a logistics detail and became a competitive feature.

The retailer logic is better still.

Pickup deletes the most expensive leg of e-commerce and converts a digital order into foot traffic. It also turns a store estate from a cost problem into a defensive asset.

Walmart captured 30.9% of US grocery e-commerce in 2025 partly because 90% of the American population lives within ten miles of one of its stores, which puts same-day fulfillment within reach of 93% of households. For any retailer competing with a pure-play, proximity is the one thing that cannot be copied quickly.

Kroger’s $2.6 billion lesson

Which brings us back to the warehouses.

Kroger and Ocado announced their partnership in 2018 with plans for as many as twenty automated customer fulfillment centers. The rollout paused in 2023. A site-by-site review followed in 2025.

Then came the decision to close the centers in Wisconsin, Maryland, and Florida, book $2.6 billion in impairment charges, and chase roughly $400 million of e-commerce profitability improvement in 2026 by doing the opposite of what the original plan required.

The new approach was pilot capital-light, store-based automation in high-volume markets, make Instacart the primary delivery provider, run DoorDash delivery from around 2,700 stores, and expand with Uber. A planned center in Charlotte was cancelled outright, along with a spoke facility in Nashville.

It would be easy to read this as an automation failure. It was not. The robotics worked. What failed was the arithmetic underneath them.

A large centralized automated facility only pays back if enough orders flow through it, and US grocery demand is spread thin across a lot of geography. Kroger built for a density that never showed up, while its customers quietly chose the parking lot instead. Meanwhile Walmart, running the same era with a store-first approach, now delivers 35% of store-fulfilled orders in under three hours and is retrofitting automation into distribution centers that were already busy.

Dark stores work. Just not everywhere.

If density is the deciding variable, then the dark store question has an obvious answer: it depends entirely on the city.

India is the proof that the model can work. There are more than 6,000 dark stores operating nationally, and Bernstein’s analysis found over 3,800 of them clustered in just the top eight cities, where population density delivers enough orders per store to make the unit economics land. In smaller towns the same store can take six to twelve months to stabilize.

Same technology, same playbook, completely different outcome, decided by how many people live within a few kilometres.

Europe ran the experiment the other way. Over $18bn of venture capital poured into speedy grocery in 2021, and Getir alone was briefly valued near $12bn before retreating to Turkey, taking Gorillas down with it. The delivery promise was real. The order density needed to fund riders, rent, and refrigeration was not.

France added a lesson nobody had modelled. When authorities reclassified dark stores as warehouses rather than shops, the zoning change helped push the remaining operators into liquidation. Planning law is a fulfillment risk, and it belongs on the spreadsheet next to labour and rent.

Pickup is not free either

None of this makes click-and-collect the safe default. It makes it the model whose costs are hidden inside store operations instead of a capex line.

Retailers currently lose an average of 6.4% of gross sales to in-store operational failures, up from 5.5% a year earlier, with out-of-stocks the sharpest-rising problem. That number matters more for pickup than for anything else, because an order promised online and unavailable at collection damages trust far more than an empty shelf a shopper discovers themselves. One is a disappointment. The other is a broken commitment, with a wasted trip attached.

So inventory accuracy is not an upgrade to a pickup programme. It is the precondition. Retailers that sequenced it correctly, using shelf-level visibility to build digital twins, improved picking efficiency by roughly 40% year over year. Retailers that launched pickup on top of unreliable stock data mostly bought themselves a new way to disappoint people. Add labour allocation, staging space, and peak-hour congestion, and the free last mile starts to look merely cheap rather than free.

And the pressure is not letting up. Amazon has been testing delivery in half an hour or less, while Sam’s Club reported roughly 65,000 one-hour deliveries within three weeks of launching. Pickup won on economics, not on speed, and the speed floor keeps dropping.

How to actually choose in 2026

The practical takeaway are these:

1. Measure orders per catchment before committing capital. Not total e-commerce growth, not national trend lines. Orders inside a delivery radius. That number tells you whether a dark store or an automated center can ever pay back, and it is knowable before you build.

2. Start with the assets you already own. Stores are paid for. Using them as fulfillment nodes is the cheapest capacity available, and it is reversible in a way that a warehouse lease is not.

3. Rent the last mile before you build it. Kroger’s pivot to Instacart, DoorDash, and Uber is not an admission of defeat. It is a decision to pay variable costs until volume justifies fixed ones.

4. Automate where volume already exists. Automation multiplies throughput that is already there. It does not create demand, and it punishes optimistic forecasts harder than manual operations do.

Walmart gave away the whole thesis in a naming decision. It used to call its automated store-attached facilities market fulfillment centers. It now calls them Accelerated Pickup and Delivery. The technology did not change. The admission of what customers actually want did.

Density is the strategy

The fulfillment debate has been framed as a contest between models, as though one of them is the future and the rest are transitional. That framing sold a lot of warehouses.

The better framing is that BOPIS, dark stores, and micro-fulfillment are all correct answers to different density conditions, and the job is matching them honestly rather than picking the most advanced one.

Dense urban catchment with high frequency and small baskets: dark stores earn their rent. Suburban catchment with large planned baskets and a store nearby: pickup wins, and it is not close. High-volume market where pickup demand already strains the aisles: that is where automation finally makes sense, in the store rather than instead of it.

Kroger spent $2.6 billion learning that the model has to fit the map. That lesson is now available for free, which makes it the cheapest thing in fulfillment this year.

Count the orders in the radius first. The model will pick itself.