Onboarding & training: support grounded in your procedures

I answer with the procedure, its date, and the exact passage — or I say I do not know.
On a common question: "the procedure of 14/03/2026, point 4.2, says: 'the case is passed to the handling department within three working days of the last document being received'." The date matters as much as the answer — a 2019 procedure and a March one are not quoted the same way.
When two documents say different things: I give both, with their dates, and I say which is more recent without saying which applies. A recent procedure does not always repeal an older one, and deciding on the department's behalf would be exactly the kind of service one does not render.
When nothing establishes it: I say so. Across the quarter's 612 questions, I answered "I don't know" 71 times — and those 71 are the most useful part of my work.
Why: a plausible unsourced answer becomes a practice. A newcomer applies it, teaches it in turn, and three years later it is the rule — with nobody able to say where it came from. 612-questions_71-unanswered.pdf541 sourced · 71 unanswered
⛓ Source · 612 questions this quarter, the authority's dated procedures

They fall into three families, and two are easily dealt with.
· 38 concern a procedure that exists but that I do not hold. It is on a share you have not opened to me, or in a paper binder. That is not a gap in the authority, it is an access to grant.
· 21 concern particular cases calling for a judgement each time. There is nothing to write — the right answer is "ask your manager", and that is what I say.
· 12 concern procedures that do not exist, including the question asked 34 times. Those are the only ones needing writing work, and they can be counted on your fingers.
What that gives you: a list of 12 procedures to write, ranked by how often they were missing. It is not a training plan, it is a shopping list — and it is short.
What I publish about the 71, and why it is not a failure rate: they are this page's yield — 38 accesses to grant, 21 judgements that need never be written down, 12 procedures to write. A tool that hid its unanswered questions to improve its answer rate would render the opposite service to the one it is asked for: it would manufacture doctrine instead of pointing at a gap. 71-unanswered_3-families.pdf38 access · 21 judgements · 12 to write
✎ Support · 12 procedures to write, ranked by frequency

What is done is done; what matters is the 15 September session.
What I record: the pack describes a four-step circuit. The procedure of 02/04 describes three — an intermediate approval was removed.
What that produced: two sessions, 23 officers trained on a circuit that no longer exists. They apply an approval nobody is waiting for, and it holds cases up on average two days.
What I checked before telling you: I looked at the cases handled by those 23 officers since their session. The two-day hold-up shows up there, and not on the others.
What I attribute the hold-up to, and therefore where I send it: to the pack, not to the 23 officers — they apply what they were taught, and a named list adds nothing: the information to pass on is the same for all of them, and it goes through the trainer.
What I propose: correct the pack before 15 September, and a two-line message to the people trained on that circuit, without distinguishing who applied what. out-of-date-pack_2-sessions.pdf4 steps vs 3 · 23 officers · 2 days of delay
⛓ Source · training pack, procedure of 02/04/2026, cases handled since the 2 sessions

It is not ignored: it cannot be found, and that is not the same thing.
What I record: the procedure was updated on 11/06. Zero views since. Over the same period, the old version is opened 4 times a week — from attachments to emails of 2024 and 2025.
What that means: officers do have the procedure, but the one they have is the one somebody once sent them. Nobody goes back to the source, because an attachment is faster to find again than to search for.
What it actually comes down to, checked: officers apply the document they hold, which is exactly the behaviour asked of them — nobody told them another one existed, and an attachment is faster to find again than the source.
What I propose: two simple things. Tell anyone opening an out-of-date version that a more recent one exists, with both dates. And, when an update is published, notify those who opened the old one in the last twelve months — without keeping who, only by sending. up-to-date-procedure_0-views.pdf0 vs 4/week · the old one circulates as an attachment
✎ Proposal · flagged on opening — no list of readers kept

I compose exercises on your own content — your procedures, not a generic catalogue.
What I build: from each procedure, questions and their expected answers, written in the procedure's own words and tied to its exact paragraph. Across your 68 procedures, that gives 412 questions: questions of fact — what deadline, what document, which department — questions of situation — “the user brings an expired document, what do you do?” — and questions of overlap, where two procedures meet.
What exercises on your content make visible: at the first session, 3 questions out of 412 had an answer in none of your documents. That is not a gap in the staff: it is a procedure that was never written — exactly the one behind the 34 questions asked in three months with no documented answer. The exercise acted as a revealer, and the gap is named instead of being carried by the staff.
What is measured, and what is not: the result of an exercise goes back to the person who took it, and to the team aggregated by procedure — “the deregistration procedure is failed by half the department, and always at the same step”. No individual result goes up to management, and the reason is mechanical: a training score that goes up becomes a mark, the mark leads to revising the exercise instead of the procedure, and the instrument stops telling you where the text is obscure. What needs correcting is the procedure that trips up half the department.
What that gives back: writing a set of exercises on one procedure took your training lead 2 hours; reading over a proposed set takes 20 minutes. Across 68 procedures revised once a year, more than 113 hours given back, rounded down.
The figure that does not flatter me: of my first 412 questions, 58 were ambiguous — 14 %: two defensible answers, and the staff were right to challenge them. All of them concerned procedures with two coexisting versions in your document library — including June's, which nobody opens while the old one circulates as an attachment four times a week. I now compose from the version in force alone, and I decline to compose on a procedure that has two: the list of duplicates is handed to you, and I have stopped building exercises on an uncertain text.
⛓ Sourced · 412 questions drawn from 68 procedures, 3 with no documented answer, 14 % ambiguous and their cause

Not yet, and this is the right moment to say so.
What I record: three arrivals on 01/09, induction path not open. Across your last five arrivals it was opened between 9 and 15 days ahead, 12 on average. There are 24 days left: you are ahead of your own practice.
Why I flag it anyway: the three arrive on the same day, and two are in the same department. Across your previous arrivals, only once did two people arrive together — and that path had been opened 15 days ahead, the maximum in the series.
What I prepared: the three paths, from the job descriptions and from what appeared in the paths for equivalent posts. Each step carries the document that documents it, and steps with no document are flagged as such — there are two out of eleven.
What is left to do, and I have reduced it to one send: the invitations are drafted, slot by slot, for the eleven steps and the people doing the welcoming, with the duration and the document for each step. They are sent and accepted by the people themselves — an induction path assumes people are available, and that has to be asked. One word from you and the three paths for 1 September open today: 24 days ahead, against 12 as a rule. 3-paths_ready.pdf11 steps · 9 documented · 2 flagged
⛓ Source · last 5 arrivals, job descriptions, paths for equivalent posts

The questions, never their authors — and it is the rule that holds up this whole page.
What is kept: the question asked, reworded without what identifies it, the answer given, the procedure cited with its date, and the fact that there was no answer.
What is not kept: who asked the question, how many questions a person asked, which procedures they consulted, and how long they took to find them. None of that is available to a manager, because none of it exists.
Why, and it is the most important point on this page: a newcomer who knows their questions are being counted stops asking. They guess, they ask a colleague, or they apply what they believe. The tool would then have destroyed exactly what it was meant to produce, and the damage would be invisible — nobody measures the questions that were never asked.
What the monthly summary contains: the most requested subjects, the questions with no written answer, and out-of-date documents in circulation. Three indicators about your documents, none about your officers. what-is-kept.pdf5 items kept · 4 impossible
✎ Framework · retention periods to be set by the authority