Right, but to give one obvious example, you just ruled out a whole industry of trainers who charge a high day rate for giving courses in specialised fields. You only know what bringing those people in was worth after your newly trained team gets better results.

Even then, unless the training is in some convenient sales/marketing field where the resulting conversions are easy to quantify, you may not be able to put a dollar amount on the long term value of the knowledge and insights your people gained from the training for a long time.

As an extreme example, consider the possibility of a one-day training course for software developers taught by a world class expert for $10K. Maybe one of your developers learned a new defensive programming technique five hours into that day, and that technique later prevented a bug that would have cost your business $1M to rectify. Obviously that training had a very high return on investment, but how is the trainer supposed to demonstrate that to his next client?

Sales collateral of very highly-paid trainers should have evidence of value and a great way to do that is exactly those kind of real examples, which a highly experienced and effective trainer will have picked up over the years. Reputation is secondary evidence of value of course, though considerably weaker. Strong evidence of value is how people can be persuaded that it's a sweet deal to pay huge amounts to consultants in any field where those consultants come with a long list of similar organisations they helped.

For the rest of the training market it would come down to a client requirement for the service, general qualification to do the job, and typical market price - existing budgets and product pricing kicks in at this point rather than ROI on specific individuals or teams. We see this in the dev contract market as well and that's what the article is about - commodity resource vs stars.

tl;dr if you can't demonstrate exceptional value, you will trend towards commodity.