[{"data":1,"prerenderedAt":19},["ShallowReactive",2],{"post-admin-is-not-a-neutral-department":3},{"slug":4,"title":5,"seoTitle":6,"excerpt":7,"content":8,"tags":9,"readingTime":16,"publishedAt":17,"published":18},"admin-is-not-a-neutral-department","Admin is not a neutral department, or: what one gigabit port costs twenty people","Admin Is Not a Neutral Department","For a long time I held a quiet, unexamined view of the administrative department — a place where not much value is made, and not much is lost. Two small things at my current company have taken that view apart. The department with the widest reach inside a company is also the one with the weakest feedback loop, and a bad one does not fail neutrally. It fails onto everyone else.","\n## The observation\n\nFor a long time I held a quiet and unexamined view of the administrative department. Not a place where much value gets made. Not a place, I assumed, where much gets lost either. Low ceiling, and a floor sitting just underneath it. If the person in the chair was mediocre, the cost of the mediocrity would be small, because the surface the job touches is small.\n\nI have come to think the floor is a great deal lower than I assumed, and that the surface is not small at all. Two things at my current company have taken the old view apart, and I have been turning them over for a few weeks without landing anywhere comfortable.\n\n## Two examples, both small\n\nThe office network has been bad for as long as I have worked here. People were called. Causes were proposed. Technicians came and went, and nothing measurable changed. A few weeks ago I stopped guessing and looked at the hardware, and the answer was sitting on the wall: the router in the rack has gigabit ports. The line the company leases runs somewhere between three and four gigabits.\n\nSo we pay for three to four thousand megabits and pass less than a quarter of that through the box everything has to cross. Twenty-odd people spend their working day behind that quarter. The remedy is a few hundred yuan of hardware and an afternoon of somebody's attention. No engineering difficulty, no vendor negotiation, no budget cycle. It is one purchase.\n\nI took it to admin the same day, with the diagnosis and the fix attached, so that nothing was left for them to work out. They were surprised in the moment. They mentioned it to the HR colleague at the next desk, who was also surprised. Two months later the office runs at exactly the same speed, and as far as I can tell nobody has looked at the router since I did.\n\nThe second example is older. Three or four months ago, a teacher in the training centre had a machine that would not behave, and admin found somebody to come and repair it — sourced off some platform, I never learned which one. During the repair the machine picked up an infection. The teacher's spreadsheets came back encrypted. The original fault was never fixed. Eventually there was nothing to do but let the man go and absorb the loss.\n\nToday I watched the same admin colleague call the same man back.\n\nI do not read this as malice, and I am fairly sure it is not carelessness in the ordinary sense of the word. From where he sits, the task was to find someone who would come to the office. That task was completed last time. A completed task, inside a system that keeps records of completed tasks, is a good precedent — the number was still in the phone, the man had answered before, the box on the sheet had been ticked. Whether the machine got fixed, and what the repair cost the teacher, belongs to a column he is not responsible for.\n\n## What the two have in common\n\nBoth are the gap between a task and an outcome, and both show what happens when an organisation quietly decides that the task is the thing it will keep records of.\n\nThe ticket closes when somebody shows up. It does not close when the problem goes away. Once that substitution is in place, every incentive points the same direction: find a body, book the visit, mark it done, move on. Calling a better technician is more work than calling the one already in the phone, and the extra work buys nothing the system can see. Buying a new router is more work than nodding at the person who raised it, and the extra work buys nothing the system can see either.\n\nThis is the same failure I wrote about in [[why-kpi-deserves-serious-skepticism|the KPI piece]], arriving from a different direction. The proxy is not merely an imperfect picture of the work. Given enough time the proxy becomes the work, and everything the proxy cannot see turns into a favour somebody is being asked to do for free.\n\nA question follows from that, and I have watched it form on enough faces to be confident it is close to universal. When every part of a job has been written down and quantified and put on a sheet of paper, and something arrives that is not on the sheet, the person receiving it asks: what do I get out of doing this. It is not a cynical question. It is the question the system has trained them to ask, and having trained them to ask it, the system does not get to be indignant about the answer.\n\nI made a related argument about the individual level in [[what-the-two-image-rule-misunderstands|the piece on what the two-image rule misunderstands]] — that a reward introduced over an activity tends to crowd out the reason a person was doing it, rather than stack on top of that reason. The version I am watching in admin is worse in one respect. In a creative team, what the carrot eats is the joy. In an administrative function, what it eats is the initiative to notice that something is broken and to keep pushing until it is not.\n\n## The cost has no owner\n\nThe second mechanism is the one I keep underestimating, and it is the reason I was wrong about the ceiling and the floor.\n\nAdmin failures are not paid for by admin. They are paid for by everyone else, in small amounts, on a ledger nobody consolidates. Twenty-odd people work a little slower for two months. A teacher loses a week of files and rather more than a fortnight of trust in her own equipment. None of that gets attributed. If you asked the company what the network problem has cost, the honest answer is that nobody has any idea, and the reason nobody has any idea is that the invoice was never issued to a single desk.\n\nMeanwhile the thing that would have fixed it costs a few hundred yuan. That asymmetry is the whole argument. A department whose failures are cheap to repair and expensive to leave alone, and whose costs land on other people's ledgers, is not a low-stakes department. It is one where the stakes are unusually high and unusually well hidden.\n\n> An administrative function is not a low-variance seat. It is a high-variance one that nobody is watching. The reach is the entire company, the feedback belongs to no one, and a bad appointment does not fail neutrally — it fails quietly, onto twenty other people's work.\n\n## The principle, stated tentatively\n\nThe departments that touch everybody need better people than their job descriptions imply, precisely because nothing in the structure will tell you when they are failing.\n\nI want to be careful with that sentence, because it is easy to hear as a complaint about individuals, and it is not one. The person who called the same technician twice did the job he was given. The failure is that the job he was given was to find a body rather than to make a problem go away, and nobody in the structure was positioned to notice the difference.\n\n## What I have come to think, for the company I would like to run\n\nI do not have this worked out, and some of it will not survive contact with an actual payroll. But I have stopped treating the administrative hire as a generic seat, and I know roughly what I would try instead.\n\nHire for judgment rather than for compliance. What the job actually requires is somebody who will look at a router, or at a repair history, and form a view. That is a taste question, and taste questions do not get solved by writing a longer task description.\n\nGive the function a memory that does not live inside one person's phone. Who was called, what they did, what it cost, whether it worked. A vendor list that remembers is a system, and the [[why-dirty-work-deserves-serious-consideration|dirty work principle]] applies here as cleanly as anywhere: the record belongs to the system, the judgment about who to call next belongs to a person. What happened with that repair is not that somebody forgot. It is that nothing was in place to remember on his behalf.\n\nMeasure the state of the office instead of the count of the tickets. Walk the floor. Ask three people how the wifi is. This is the same move I keep arriving at everywhere else — when the proxy is failing, the fix is not a better proxy, it is going and looking.\n\nClose the loop with whoever raised the thing. Two months of silence taught me something quite specific: not that admin is slow, but that reporting a problem here does not produce a response, and that I should therefore stop reporting and start routing around. That is an expensive lesson for a company to teach its staff, and it costs almost nothing to avoid.\n\n## What I am not saying\n\nI am not saying the people in these roles are lazy or unintelligent. Both failures I described are what a reasonable person does inside a structure built the way this one is built.\n\nI am not saying every office needs a lavish network. The point is not the specification; the point is that we pay for one thing and receive a quarter of it, and that this stayed invisible for months.\n\nI am not saying the answer is a sharper set of administrative KPIs. That is the trap I have spent several pieces arguing against, and administrative work — full of small unrecorded judgments about who to call and what to check — is close to the worst available candidate for piecework.\n\nNor am I saying the description I just gave solves the hiring problem. Judgment shows up in an interview only when the interview is built to make it show up, and most interviews are not.\n\n## Where this connects\n\nThis sits underneath [[why-kpi-deserves-serious-skepticism|the KPI argument]] and next to [[why-dirty-work-deserves-serious-consideration|the dirty work argument]], in the same family as [[the-two-image-week|the two-image week]]. The through-line is the one I keep circling: a system that can only see the form will eventually get the form, and the work goes quietly missing behind it.\n\nThere is also a link to something I noticed a while back and filed under [[professional-matters-to-professional-people|professional matters to professional people]]. Support functions fail in a characteristic way — they buy or arrange the thing that looks like the answer, and the people who have to live with it route around it. What I would add now is that the failure is not confined to expensive systems nobody uses. It shows up just as clearly in a three-hundred-yuan router that nobody buys.\n\n## What the literature already says\n\n**Eliyahu Goldratt, *The Goal*.** The router is a constraint in the most literal sense Goldratt meant it — every packet in the building crosses that one device, so its capacity is the building's capacity, and every improvement made anywhere else stays invisible until that one is lifted. What his argument adds, in my reading, is the harder half: attention rarely finds the constraint on its own, because the constraint tends to sit somewhere unglamorous that nobody owns.\n\n**Henry Mintzberg, *The Structuring of Organizations*.** Mintzberg separates the operating core from the support staff arranged around it, and he is clear that the support side is structurally hard to evaluate, because its contribution shows up as the absence of friction elsewhere. An administrative function doing its job well produces nothing you can point at. That is exactly why the job attracts proxies, and exactly why the proxies mislead.\n\n**Albert Hirschman, *Exit, Voice, and Loyalty*.** Hirschman's argument is that people respond to decline either by leaving or by speaking up, and that voice survives only where it produces a response. I used voice once, in the most convenient form available — the diagnosis and the fix, handed over on the same day. Nothing came back. What I did afterwards is what his framework predicts: I stopped raising it and started working around it. The router is still there, and the more durable damage may be that one more person in the building has learned there is no point.\n\n## A postscript\n\nThe reproach in that last paragraph is aimed at me as much as anyone. I raised it once and went back to my own work, and I have been telling myself since that raising it once was enough, which it evidently was not. Voice used a single time is not really voice. It is a receipt you keep so that you feel entitled to the complaint later.\n\nThe other half of this is the human resources function, which I have been watching with the same slow change of mind and which deserves its own piece rather than a paragraph at the end of this one. The failures rhyme, but the reach is different, and the reach is what makes it worse.\n    ",[10,11,12,13,14,15],"Org Design","Process","KPI","Dirty work","Team Design","Management",10,"2026-09-01",true,1788356310896]