Policy questions: when the written rule and the practised rule differ

The question is a symptom of two sources

When the same question is asked every week, it is almost never because the answer is hard. It is because there are two answers in circulation: the handbook from four years ago and the way the team has done it since. People ask a colleague rather than read, because reading has burned them before. The fix is to decide which source is authoritative and retire the other - and that decision is a management act, not a software one.

Three kinds of policy question, three correct behaviours

A settled procedural question - how to claim expenses, how much notice leave needs - has one published answer and should be answered. A question with conditions - whether this specific case qualifies - should state the rule and the conditions and stop short of the verdict. A question with legal or employment consequence - dismissal, grievance, pay, anything about a named colleague - must be routed to a person without an attempt at an answer. Collapsing the three into one is the usual failure.

A confident wrong answer is worse than silence

Someone who gets no answer asks again. Someone who gets a plausible wrong answer acts on it, and nobody discovers the error until it has consequences. That asymmetry is why the correct behaviour on an unanswerable question is to say what is not known and name who owns it - and why "did it answer everything?" is the wrong question to measure, while "was anything answered wrongly?" is the right one.

The log is the inventory of what to rewrite

The list of what people actually ask, in their own words and at the frequency they ask it, is the most useful document an operations team can have and almost nobody has it - because the questions arrive verbally, at desks. Collected for a month, it names the five handbook sections worth rewriting, and it does so in the words staff actually use rather than the words the handbook uses.