What Coles dropping Palantir means for AU retailers

Inside Retail reported on 7 September 2026 that Coles has confirmed its three-year data-analytics agreement with Palantir Technologies will not be renewed when it expires in 2027. The decision follows a GetUp campaign that gathered more than 85,000 petition signatures objecting to what it called military-grade technology being used on supermarket shoppers. A Coles spokesperson told Inside Retail the platform "has delivered value across Coles' operations, particularly in rostering, store operations and supply chain planning", that it is "an operational platform focused on helping stores run more effectively", and that "Palantir does not control the data, nor can it independently repurpose it". Coles did not answer Inside Retail's question about why it is not renewing. Nothing about this has appeared in an ASX release, so treat the confirmation as reported by trade press, not filed.

Why it matters here

Coles is not unusual in what it was doing. Its own 2026 Annual Report says management expanded "the use of advanced analytics and digital tools to improve forecasting, rostering, store-specific ranging, operational decision-making and team member productivity". Woolworths Group's F26 results describe "AI-driven workforce management and rostering capability". Every retailer with more than a few hundred stores is doing some version of this. What was unusual about Coles was that the vendor's name carried a political charge in Australia, and that Coles waited until a petition had 85,000 signatures before explaining, in plain language, what the software did.

The take

I think Coles made the wrong call for the right reason. The right reason is that a supermarket's licence to hold data about its customers and its staff depends on people believing it will use that data sensibly. If a vendor relationship is corroding that belief, the relationship costs more than the contract. The wrong call is treating the vendor as the problem. Whatever replaces Palantir will do the same things: forecast demand, roster people, plan the supply chain. If Coles cannot explain those uses well now, it will not be able to explain them under a different logo either.

The lesson for the rest of us is that the explanation has to exist before the campaign does. Rostering software that uses sales forecasts to decide how many people work Saturday afternoon is defensible. "Military-grade analytics on shoppers" is not, and once the second description is in circulation the first one sounds like spin. Coles' statement, that the platform is operational, that the vendor cannot repurpose the data, that it is about products on shelves and team members in the right place, is a good statement. It needed to be on the corporate site in 2024, when the contract was signed, not in a trade press quote in 2026.

There is a second, quieter lesson about switching cost. Coles has given itself until 2027 to move rostering, store operations and supply chain planning off a platform that was, by its own account, delivering value. That is a real programme with real risk to store execution, undertaken to solve a reputational problem. Retailers choosing analytics platforms now should price that risk in: how portable is the model logic, how portable is the data, and how loud would it be if you had to leave.

What to watch

Watch whether Coles names a replacement, and whether the FY27 filings show any change in the forecasting and rostering language that the FY26 report leaned on. Watch whether Woolworths, Wesfarmers or Metcash are asked the same question about their own vendors. And if you run a retailer, take an afternoon this month to write down, in one page, what your operational analytics actually do with data about staff and customers, in words a customer would accept. If you cannot, that is the finding.