← AI in physical storesC2 · LESSON 7 OF 7

C2, Lesson 7: Click-and-Collect, Store-Based Fulfillment and the Roadmap

Tags: Sourced / General / Our view / Illustrative. About 13 minutes.

1. The story: Kroger closes its robot warehouses, November 2025

On November 18, 2025, Kroger said it would close three automated fulfillment centers, in Pleasant Prairie, Wisconsin; Frederick, Maryland; and Groveland, Florida, in January, and would monitor its five remaining facilities. Kroger had partnered with Ocado in 2018 to build warehouses where robots pick and pack grocery delivery orders. The plan was 20 locations; eight had been built. Kroger said it would take a $2.6 billion charge in its fiscal third quarter and expected the closures to improve e-commerce operating profit by $400 million in 2026. [Sourced: AP, Nov 18, 2025; Reuters, Nov 18, 2025] CEO Ron Sargent said that in most locations it makes sense to use stores to fill delivery orders: stores are closer to customers, so orders can be delivered more quickly and cheaply, and Kroger could deliver in under two hours from 97% of its 2,700 stores. He added that in some high-density areas the automated sites perform better. [Sourced: AP, quoting Kroger's September 2025 investor call] Kroger was also expanding partnerships with DoorDash, Uber Eats and Instacart. [Sourced: AP] An eMarketer analyst told Reuters that the pandemic bump did not hold and shoppers moved back to stores. [Sourced: Reuters; an analyst's view] The lesson: the store is also a warehouse. The question is not "robots or stores" but how many orders per week, how dense, and what fixed cost. [Our view]

2. The one idea

Click-and-collect (shoppers order online and pick up at the store) uses the store's shelves and staff as the fulfillment center. AI helps in three places: choosing substitutes when an item is out, sequencing picks through the store, and forecasting pickup volume to staff it. The key economics: fixed-cost warehouses need volume; store picking scales with orders. [General]

3. The kitchen-table version

You can fill online orders from your own shelves (no extra rent, your own time) or rent a separate warehouse and hire a team there. The warehouse is faster per order but costs rent whether or not orders show up. [General]

4. How store fulfillment works

  • Order intake with live stock data (links to Lesson 2: shelf accuracy decides substitutions and cancellations).
  • Picking by associates with a handheld; AI sequences the route and suggests substitutes. Walmart describes LED "Pick to Light" labels for this (Lesson 5). [Sourced: Walmart post, Mar 2, 2026]
  • Handoff: curbside or locker.
  • Delivery by the retailer or a third party.
  • Forecast: how many orders at which hours, so labor fits (Lesson 6). [General]

5. Worked example: when does a warehouse beat store picking?

All numbers [Illustrative], computed by script. Basket $95 at 25% gross margin: $23.75 gross profit. Store pickup: a picker takes 35 minutes at $20 an hour = $11.67, plus $2.00 handling. Contribution: $10.08 per order. Home delivery from the store: add a $9 driver cost, a $5 customer fee: $8.08. Automated warehouse: picking costs only $3.00 per order, but fixed costs are $40,000 a week. Contribution per order with delivery: $23.75 - 3 - 9 + 5 - fixed per order.

Orders a weekWarehouse contribution per order
2,000-$3.25
4,000$6.75
6,000$10.08
8,000$11.75
Break-even is about 2,400 orders a week; it only beats store delivery ($8.08) above about 5,000 orders a week. Below that, stores win. This is the shape of the trade-off Kroger's CEO described: dense, high-demand areas can justify automation, most places cannot. [Our view] Do not copy these numbers; use your own.

6. Costs, risks and what it breaks

  • Out-of-stocks and substitutions: a pickup shopper who gets the wrong item is a lost customer (Lesson 2).
  • Store congestion: pickers and shoppers share aisles.
  • Margin: free pickup is a service you pay for; track the contribution per order, not just sales.
  • Third parties: delivery apps bring reach and take a share and customer data. [General]
  • Stranded cost: fixed-cost assets can turn into a multi-billion-dollar write-off, as at Kroger. [Sourced: AP; Reuters]

7. The course roadmap: what we saw across seven lessons

  1. Data first: shelf and stock accuracy (L1-L2) feed everything else.
  2. Checkout and loss (L3-L4): measure the loss you add; an alert is a guess.
  3. Labels and labor (L5-L6): speed is easy; trust and predictability are the work.
  4. Fulfillment (L7): the store is a node in a network; volume decides. Pattern from the openers: Walmart ended its shelf-scanning robot (Nov 2020), Amazon pulled Just Walk Out from Fresh (Apr 2024), Dollar General reversed self-checkout (Dec 2023), Kroger closed three robot warehouses (Nov 2025). Each time a more automated design gave way to a simpler one in some settings. That does not make automation wrong; it means test small and measure. [Our view]

8. Monday checklist

  1. Record your pickup volume by hour and your pick time per order.
  2. Compute contribution per order: gross profit - labor - handling.
  3. Track substitution and cancel rates.
  4. Decide a max pickup volume per hour your store can handle.
  5. For any fixed-cost proposal, compute the break-even orders per week.

9. Try it yourself (no code, 30 minutes)

Use Section 5 with your numbers. Find the weekly order volume at which a fixed-cost option beats picking from your own store.

10. Self-check

  1. What did Kroger announce on November 18, 2025? Closing three automated fulfillment centers, $2.6 billion charge.
  2. Why did the CEO prefer stores for most places? Stores are closer, so delivery is faster and cheaper.
  3. In the example, at what volume does the warehouse start to beat store delivery? Above about 5,000 orders a week.
  4. Why do substitutions matter? A wrong substitute can lose the customer.
  5. What pattern shows up across the openers? A more automated design gave way to a simpler one in some settings.

11. Where this is going

If shopping agents (C1 Lesson 7) place orders, a store that offers reliable pickup windows and accurate stock will be easier to buy from. Robots at scale and store-based fulfillment will probably coexist, split by density. [Our view]

Sources

Review flags

  • Read AP in full and the Reuters lead; the Sargent quotes are AP's paraphrase of an investor call, not read from a transcript.
  • The $400 million is Kroger's expectation, a company claim.
  • RESOLVED (Oct 7, 2026): the Dec 5, 2025 Reuters piece's lead reads: Ocado will receive a $350 million one-off payment from Kroger after Kroger opted to shut three robotic warehouses and scrap another facility (search excerpt of the Reuters article; Ocado's own Dec 5 announcement is on investegate.co.uk). The lesson does not state the figure, so nothing needs changing.
  • The roadmap pattern relies on lessons 2-4 (Walmart and Bossa Nova; Amazon; Dollar General); each is sourced in its lesson.
  • Example numbers are illustrative.

Claim ledger (C2 Lesson 7)

#ClaimSourceVerifiedStatusRe-check
1Kroger closing 3 centers (WI, MD, FL) in January; 5 remain monitoredAP Nov 18, 20252026-10-06Read2027-01
2$2.6B charge; $400M 2026 profit improvement expected; 20 planned, 8 builtAP; Reuters2026-10-06Read; company claim2027-01
3Stores faster and cheaper; under 2 hours from 97% of 2,700 storesAP (Sargent, Sept call)2026-10-06Read; paraphrasetranscript
4DoorDash, Uber Eats, Instacart partnershipsAP2026-10-06Readnone
5Analyst quote on pandemic bumpReuters2026-10-06Readnone
6Fulfillment economicsOur computation2026-10-06Illustrative, script-checkednone