There is a particular kind of manager I have come to recognize, and I want to describe them before I say anything about documentation, because the description is really the whole argument. They have done everything they were told to do. For six months they have written down every late arrival and every missed deadline, every sharp exchange in the hallway and every meeting that drifted somewhere it shouldn't have, and they have done it diligently and in good faith, out of a sincere belief that the way you protect yourself and your team is to get it all down on paper. Then a grievance is filed, or a complaint lands on somebody's desk, and the file they built with such care is read for the first time by a person who was never in the room — a union rep, an investigator, eventually perhaps an arbitrator — and somewhere in the first few pages it quietly begins to work against them.
The volume they trusted turns out to be the thing that sinks them.
I have seen this often enough to stop treating it as bad luck. Most managers over-document out of anxiety, and the bloat itself weakens the record, because a file read cold by a third party is not a diary of your diligence — it is a surface for the other side to work. The more you have written, the more contradictions there are to find, and the more moments where the tone slipped and gave away a mind already made up. A neutral is not impressed by thickness. They are reading for something narrower, and much harder to fake: was this fair, and can I follow it?
That question is increasingly what the people who decide these things are looking at. Studying three decades of harassment arbitrations in unionized workplaces, Abraham and Voos (2021) found that arbitrators weigh not just whether they believe the complaint but whether management followed a sound process — strong procedure can support firm discipline, while weak procedure pushes toward leniency even when the misconduct is real. The Canadian just-cause tradition in Brown and Beatty says much the same: what an arbitrator looks for is clear evidence of the conduct, a documented rationale for the penalty, and a traceable trail of notice and chance to respond. None of that is a request for more paper. It is a request for the right paper.
So the discipline I have landed on is not to document more. It is to build the roughly twenty percent of the file that carries the weight, and to let the rest go. That twenty percent has a through-line you could say in a single breath — what was expected, what happened, what was tried — and everything else either serves that line or muddies it.
In practice it comes down to three moves, and the craft is knowing which one a given sentence calls for. Some things you capture verbatim: the accounts in people's own words, the admissions and the denials, and the procedural facts that later become load-bearing — notice given, representation offered and attended, timelines respected. Some things you summarize, and neutrally, in the register of weight rather than character: "more consistent with," "not corroborated by," "the evidence did not support." And some things you leave out, because they were never going to help and were always going to cost.
I think of a case — a composite of several I have handled, none of them nameable here — that has become my private measure of what good looks like. It was a harassment file of the hardest kind: fragile credibility on both sides, no clean corroboration, a genuine he-said, she-said. Nothing about the facts was going to rescue it; what held the record together was the procedure. Because the facts were messy, the process slowed rather than sped up, and every key conversation was closed with a short follow-up email confirming only what had happened procedurally — the interview took place, representation was offered, here is the next step — without ever promising an outcome. The decision, when it was finally written, kept three things in three separate rooms: the findings of fact, the procedural narrative, and the recommendation on discipline.
And what kept it clean was as much what stayed out as what went in — though not in the way that phrase usually implies. It did not mean pretending I hadn't noticed how people carried themselves; demeanour and motive are real factors in any honest credibility assessment, and I do weigh them. It meant refusing to let an impression stand in the file as if it were a finding. "She seemed hysterical" and "he was defensive" are conclusions with the reasoning stripped out, leaning on demeanour far more than demeanour can bear. When defensiveness is worth noting, the work is to ask what it is doing — minimizing, deflecting, or simply the reaction of almost anyone who has just been accused, guilty or innocent alike — and to set the innocent reading beside the adverse one. What stays out of the file is the untested label, the guess at motive dressed up as fact, and the promise about an outcome I had not yet earned the right to reach.
Underneath all of it is a principle I keep coming back to: an organization is by its nature a slow-moving beast, and its deliberateness is not a flaw to be overcome but a strength to be used. I want to be careful with the word slow, because some moments genuinely call for speed — placing someone on paid administrative leave while you look into things is often the responsible thing to do quickly rather than slowly. But once you are inside the process itself, the pull to rush it is the thing to resist. There is no room in a proper process to act covertly, and no real advantage in the element of surprise. Surprise is not strength; process is. It is the open, deliberate path — visible to everyone, moving at the organization's true speed — that makes a fair outcome close to inevitable, and makes the record hold when it is read at arm's length a year later.
If you want somewhere concrete to start this week, start with the follow-up email: after every conversation that might one day matter, send a short note confirming what happened procedurally, not what you concluded. Run each thing you are about to write through the capture-summarize-omit question first. And keep three kinds of sentences out of the file entirely — the ones that characterize a person's demeanour as though it settled the matter, the ones that guess at motive, and the ones that promise an outcome you have not yet earned.
Here is the part I did not expect when I started this work, and the reason I wanted to write about it first. The lean, fair, high-signal file — the one that survives the arbitrator — is the very same file that treats the employee fairly. The clarity that protects you is the clarity that gives them no surprises and a genuine chance to respond. In the case I keep returning to, the complainant felt fully heard even though trauma shaped how she came across, and the respondent, who did not like where it landed, could still see that he had been given notice, real chances to answer, and a path to recourse — so it felt disappointing to him, not arbitrary. Procedural fairness, it turns out, is the strange place where protecting yourself and being fair to the other person stop being two acts and become one.
What I have not solved — what I am not sure documentation can solve — is what happens to the person left standing under a "no finding," carrying a suspicion that was neither confirmed nor cleared. A clean file protects the process. It does not lift that particular shadow, and I am still working out whether anything can. That is a thread for another week.
If you have watched a thick file come apart, or a lean one hold, I would like to hear how this lands for you — reply and tell me what you saw.
Sources & Further Reading
Steven E. Abraham & Paula B. Voos, "Procedural and Distributive Justice in Sexual Harassment Arbitrations: Evolution of Decisions in the Union Context," Advances in Industrial and Labor Relations, Vol. 26 (Emerald, 2021). Analysis of union-context labour arbitration decisions, 1988–2018. Emerald Insight
Donald J. M. Brown, David M. Beatty, Adam J. Beatty & Christine E. Deacon, Canadian Labour Arbitration (Thomson Reuters, looseleaf; available in print and via LabourSource). The standard Canadian treatise on grievance arbitration, just cause, and progressive discipline. Thomson Reuters