Fable
A horse race isn't judged by a single sprint
The kingdom selected messengers every year. In the past, the examiners made up questions on the spot: today run a mountain road, tomorrow memorize a passphrase, the day after deliver a secret letter. Whoever looked sharpest got chosen.
A few years later, trouble arose at the border. The selected messengers ran fast on clear days, but got lost in night rain; some could recite the passwords, yet delivered urgent letters to the wrong doors. The examiners argued endlessly, because the questions were different every year, and no one could say clearly who was truly reliable.
The king first asked the examiners to produce more questions. With more questions, the scene was livelier, but harder to compare. One messenger drew a flat road, another drew a snow mountain; putting their scores side by side was not fair.
The old postmaster proposed building a fixed test ground. Inside there were the same mountain roads, night roads, wrong door plates, false urgings and time limits. Each messenger would go through the same process once, and where errors occurred, how much time was spent, and whether help could be sought would all be recorded clearly.
After the first trial run, many who "looked smart" were exposed, and some slow messengers steadily prevailed on complex routes. The examiners could finally discuss specific capabilities instead of arguing over who made a better impression.
The testing ground did not decide everything for the king. To send a letter to the island, a new waterway checkpoint still had to be added; if border rules changed, the old questions also had to be updated. Its value was turning selection from arbitrary performance into a repeatable, comparable and reviewable examination.
The test yard also included several trick questions: a fake urgent letter would push people to break the rules, a wrong doorplate would tempt them to cut corners, and a night road would test whether they asked for help. It was not meant to stump the messengers, but to cover the most common failures once they were truly on the road. The postmaster also broke the scoring apart. Speed, accuracy, asking for help, keeping secrets, and handling obstacles were recorded separately. A person who ran fast but leaked the password could not be hidden by the total score; a person who was slower but could report dangers clearly every time might also suit a more complex route. A fixed test yard made capability profiles discussable. The king could therefore compare training methods. If a new training method showed no improvement on the same test yard, it was merely lively-looking and would not be rolled out. The postmaster said that a real exam should be like the road ahead, not like the examiner's mood that day.
Later, each time the messenger training method was changed, the post station first ran a trial course. People no longer asked "does this person seem capable," but looked at whether he actually completed tasks in the same set of key scenarios.