I stick to it: there is no successful case of agency yet.

Last week's post on agency reached over 100,000 people. The comments were the best part - genuine practitioners, on both sides, arguing from experience. Several told me it works, and named Mercedes-Benz AG India, MINI , Volvo Cars, Austria and Sweden. Others called it a disaster and pointed to Volkswagen Group VW's retreat and Stellantis pulling back. Almost everyone agreed on one thing: implementation is where it lives or dies.

So I checked the successes properly. Each one turned out to have a condition attached that disqualifies it as proof.

MINI is not an outside case. It is BMW Group selling through BMW Group's own dealer network. And BMW just pushed its main brand's switch back to 2027 - explicitly to learn more from how MINI has gone. That is not a finished success. That is a pilot still being debugged.

Mercedes India worked, by the account of the person who ran it, because of a strong brand, a small market and few dealers. Precise, and revealing. Mercedes sold nineteen thousand cars in the whole of India last year - 0.4 per cent market share. Germany is a different planet.

Austria, one of the earliest pilots, has reportedly had unhappy dealers throughout. And in Germany, more than two years in, the rollout is still partial - Mercedes has had to bring back the regional dealer support it stripped out in 2022, because running the network from the centre needed something local that had been removed.

Every example offered as proof turns out to be a small market, a pilot, or a company selling to itself through its own established network.

Nobody has yet done it: full scale, full complexity, an outside brand with no existing network to lean on.

Which is exactly the test Xiaomi says it is about to run in Germany in 2027. If you have seen it work at that scale, anywhere, I would genuinely like to know. Thank you to everyone who argued back last week.You gave me the evidence. It held up.