Grocery retail is automating at two very different speeds, and the distinction matters for anyone signing off capex this year. Distribution centre robotics has a proven, calculable return, with a live New Zealand example to point to.
In-store robotics is still working out where the line between genuine savings and expensive experiment actually sits. Start with a distribution centre, where the numbers are strongest and the evidence is local. DHL's Auckland facility runs 41 Geek+ robots across 12,000 temperature-controlled pallet spaces, New Zealand's largest third-party logistics automation project to date. International benchmarks for micro-fulfilment centres, built on AutoStore-style cube storage and already running for chains such as Ahold Delhaize overseas and Woolworths, follow a clear volume threshold: below 10,000 orders a month, manual picking generally remains more cost-effective; above 30,000 orders a month, the automation economics become compelling. One documented overseas model puts a mid-range micro-fulfilment build at roughly a six-year payback on the lower end, US$5 million against US$70,000 a month in labour savings. New Zealand order volumes at most single-site distribution centres sit well below the threshold where that math starts to work, which is worth weighing before benchmarking local plans against offshore case studies built on much larger throughput.
That is a real number, and it is a useful one, because it tells you exactly where the volume cut-off sits before you commit capital.
In-store robotics is a different proposition, and retailers should be more sceptical of the pitch than the distribution centre case warrants. Shelf-scanning robots such as Simbe's Tally are built to catch out-of-stocks and pricing errors, addressing a real cost: Simbe itself estimates retail inventory inefficiency cost the global industry more than US$1.7 trillion in 2023, a global figure with no published New Zealand-specific breakdown. Shelf stocking alone consumes over 30 percent of store attendants' working hours, according to international research, which is exactly why the category is attracting investment, though local attendant workloads and store formats vary enough that the figure should be treated as indicative rather than exact for NZ operations.
Dexterous tasks remain genuinely unsolved regardless of market. No commercially deployed robot reliably handles fresh produce without bruising it, tomatoes, peaches and berries all deform under grip pressure, and soft-gripper research has not yet closed that gap. That means the most labour-intensive, highest-shrinkage category in a grocery store, fresh produce replenishment, is currently outside what any in-store robot can actually do unattended, anywhere in the world.
The commercial read for retail grocery is not "automate the store." It is narrower and more useful than that. Shelf-scanning and price-audit robots have a defensible, data-backed case built on measurable inventory loss, though New Zealand retailers should be running their own shrinkage and labour-hour numbers rather than importing the global figure wholesale. Cleaning and transport robots suit high-traffic, structured environments and narrow aisles, Pudu's retail units are built to operate in aisles as narrow as 70cm. Dexterous stocking and produce handling remain, honestly, years away from commercial reliability regardless of what a vendor demo suggests.
For category and operations teams weighing automation spend, the distribution centre case is close to settled, and DHL's Auckland build is proof it works here. The in-store case still needs a volume and category-specific business case built on local numbers, not a blanket automation strategy borrowed from an overseas warehouse or an offshore retail chain three times the size.
Read more articles here.
