Log the time you actually spent, not the time you think it should have taken. An honest number that
makes the estimate look wrong is worth far more than a tidy one — the whole point is to find out where the
estimate is off. Count hands-on time only: if you were waiting for a page to load or a call to connect that
counts, if you were waiting for the post it does not.
Log a task
What the log says so far
By runbook
A revised standard needs at least three samples for that runbook. One timing is an anecdote — it could have been an unusually awkward company, or a slow morning.
Everything logged
Fifty tasks tells you about the pack, not about each runbook. Across 32 runbooks, fifty entries is
under two samples each — enough to correct the overall standard with confidence, and nowhere near
enough to re-time an individual procedure. Trust the headline factor; treat the per-runbook column as a
hint until it has three or more behind it.