Grocery Automation Pays When the Catchment Fills

Key takeaway

Coles says its two automated customer fulfilment centres reached positive EBITDA in their second year of operation. That is a useful operating milestone, but it is not simply a robotics victory. The centres crossed into positive EBITDA while their catchment areas expanded, same-day delivery was added, volume grew by more than 30%, and service quality ran ahead of the wider online channel.

The mechanism matters. Grocery automation does not become economic because a robot can pick an item faster. It becomes economic when enough orders can be routed through a fixed facility, at high enough accuracy, across a dense enough delivery area, without making customers wait or forcing the store network to carry all of the peak capacity.

What's changing

On August 25, Coles reported that supermarket ecommerce sales grew 26.4% in fiscal 2026 to A$5.6 billion. Ecommerce penetration reached 13.6% for the year and 14.6% in the fourth quarter, up from 11.2% in fiscal 2025. Its two Ocado-powered customer fulfilment centres, or CFCs, delivered positive EBITDA for the year.

Coles also disclosed the operational changes behind the result. It expanded the centres' catchment areas, introduced same-day delivery, installed on-grid robotic pick arms, and added automated frame loading and bagging. Store-fulfilled click-and-collect gained capacity as same-day delivery orders in Melbourne and Sydney moved to the CFCs. Coles' fiscal 2026 results put those pieces in one system: the automated sites absorbed more delivery volume while stores retained a different set of online missions.

Ocado, which supplies the technology, added two useful but company-reported indicators. It said CFC sales grew by more than 30%, faster than Coles' total online channel, and that customer net promoter scores for the CFC service were 710 basis points above total online NPS. Those figures need to be read as supplier and retailer disclosures, not independent benchmarks. They still show why the centres' financial progress cannot be separated from customer adoption and order density. Ocado's Coles update was published alongside the retailer's results.

Why it matters

Automated grocery facilities have punishing fixed-cost economics. The grid, robots, software, building and integration costs arrive before the order volume. A lightly used centre can post impressive units-per-hour statistics and still lose money because depreciation, maintenance, supervision, energy and delivery capacity are spread over too few baskets.

Coles' result points to three thresholds that matter more than a generic automation target.

First is catchment density. Expanding the geographic area raises the addressable order pool, but only helps if routes stay efficient and delivery promises remain credible. Same-day service can increase demand and improve asset utilisation, yet it can also create expensive peaks. Operators need contribution by delivery window and postcode, not just daily centre volume.

Second is mission allocation. Coles did not declare that every online order belongs in a CFC. It moved same-day delivery volume away from stores while store-fulfilled click-and-collect retained a role. That is a portfolio decision. Stores can be economical for pickup, urgent orders and dispersed markets; automation can be stronger for planned delivery baskets in dense urban zones. Routing the wrong mission into the wrong node destroys the apparent labour saving.

Third is service quality. A fulfilment centre that improves item availability, freshness and order accuracy can lift repeat use and basket frequency. That demand effect is easy to exclude from an engineering business case, but it may be essential to the payback. The 710-basis-point NPS gap does not prove retention or profitability by itself. It does suggest the automated channel was not reaching break-even by degrading the customer promise.

There is an important limitation. Positive EBITDA is not the same as an adequate return on invested capital. Coles did not disclose CFC revenue, EBITDA dollars, depreciation, maintenance capital, delivery contribution or the original capital base. Fiscal 2026 also benefited from the absence of A$103 million in major project implementation, dual-running and transition costs recorded in fiscal 2025 across Coles' automation programs. The centres can be operationally positive while the full investment still has a long payback.

What operators should do

Build the automation case around the demand and routing system, not the equipment specification.

Start with a node-level contribution model. Separate picking labour, delivery, refunds, substitutions, customer service, payment cost and promotions by fulfilment mode. Add depreciation and sustaining capital in a second view so an EBITDA milestone cannot be mistaken for a capital return.

Measure density explicitly. Track orders per catchment, baskets per delivery route, capacity utilisation by hour, and the share of volume transferred from stores rather than newly acquired. A growing centre can look productive while merely moving cost from one node to another.

Protect optionality between stores, automated centres and third-party immediacy networks. Coles is expanding CFC same-day delivery while also using Uber Eats for rapid missions. The useful architecture is not one channel winning. It is a routing layer that assigns each basket to the lowest-cost node capable of keeping the promise.

Finally, connect service metrics to economics. NPS, on-time delivery and item availability should be joined to repeat rate, basket size, complaint cost and churn. An automated site becomes valuable when better execution creates enough retained demand to fill the asset.

Bottom line

Coles' positive CFC EBITDA is evidence that automated grocery fulfilment can cross an operating threshold. It is not evidence that automation works everywhere, or that robots alone created the return. The threshold was reached through volume, catchment expansion, mission routing and customer quality.

For grocery operators, the decision is no longer simply whether to automate. It is whether the business can assemble enough dense, repeatable demand around the asset, and whether it can keep each order in the fulfilment node where the full contribution is strongest.