Noted on 8 Sep 2026

Google recently ran an experiment most councils only get to argue about in theory: reroute a small number (under 2%) of trips away from your most congested roads, and see what happens.

Published in Nature Cities in June, a proper switchback trial, six months, ~100 segments across 10 US cities, alternating which areas got the treatment day by day. Not a simulation, not a vendor whitepaper.

The result: 2% faster on the segments actually targeted, a 0.7% cut in travel time across the whole network, and over 1,000 tonnes of CO2e avoided a year, per city. All from nudging a sliver of trips onto routes with comparable travel time but less traffic.

Most mobility discussion I’ve had with councils & operators have been about physical capacity: kerb allocation, car club bay counts, EV charging density, procurement cycles measured in years. This was a software company reallocating demand at effectively zero marginal infrastructure cost, at a scale no local authority could ever touch.

It doesn't replace kerb policy. But it's a great reminder that some of what looks like a capacity problem is actually a coordination problem, and that coordination increasingly lives inside an app many residents already have installed, not inside of a transport plan.

Curious what people think of this: is a routing tweak a legitimate lever for a local authority to consider, or does relying on it just push the harder infrastructure decisions out another year?