The series the rider keeps
A rider who has worked for Deliveroo in Edinburgh for more than five years has been writing down his monthly income, the number of orders delivered and the average fee per order since 2023. In that ledger the orders delivered per hour have stayed between 3.6 and 3.8 for three years, while the average fee per order was 3.67 pounds in 2023, 3.63 pounds in 2024, 3.51 pounds in 2025 and 3.42 pounds in the first half of 2026.[1]
What makes these figures useful is that they show, on one consistent measure, how the payment for the same work changed. The ledger does not reveal which platform inputs produced each job offer and its fee. Deliveroo says information on its algorithm and fees is available on its rider website; the major platforms also say riders earn above the national living wage while on an order, excluding time waiting for offers. Those statements do not explain the calculation behind the individual fees in this ledger.[1]
What does the side doing the calculation say?
Uber denies adjusting fares to an individual driver's behaviour and attributes the differences observed to other features of its system, such as GPS. That answer remains untested; measuring it would require a breakdown showing what fee the same offer carried for different drivers.[1]
There is research pointing the same way. Work by the University of Oxford and Columbia's business school found that Uber drivers earned substantially less an hour after the dynamic pricing algorithm arrived in 2023. The Workers' Observatory, set up by riders, ran a test in Dunfermline where a group of riders logged on together, with some refusing offers below a set rate; some riders saw their rate rise briefly, one was deactivated soon afterwards and the reason stayed unclear. The major delivery platforms say riders are not deactivated for rejecting job offers. That test does not show how the fee is set; it shows how much an outside observer can see.[1]
What would close the gap?
A document could close the argument: a breakdown a rider could set against their own ledger, showing which inputs produced the fee for an order. The same shape was set out in this column on 1 September for the medical scribe tools writing into NHS records, where the error was usually caught by the patient at the end of the chain because the review that would have tested the tool before deployment had been removed. Here too the first to notice something wrong is the person keeping their own record.[1], [2]
The class action lodged in Amsterdam this week is one possible route for such a document to surface. The drivers come from the UK and the Netherlands, and their claim is that the system setting pay does not meet data protection rules. The testable outcome is a breakdown of how an order's fee was set entering the file, rather than an aggregate summary.[1]