The Invisible Work Problem: Why Quiet Months Look Like Idle Months
Somewhere in your organisation there is a spreadsheet of what IT cost this year. Nowhere in it is a line for the disk you expanded in March before it filled. The outage that would have followed did not happen, so it has no ticket, no incident number, and no memory. The invoice for the monitoring software, on the other hand, arrived on time and was filed.
This is the invisible work problem. The best months of your working life are the ones where you caught everything early and nothing fell over. From the outside, those months look like nothing happened. I spent years being mildly offended by this before I understood that it is not a judgement. It is an information gap, and monitoring data can close it. That is because the monitoring system was watching every one of those quiet mornings and wrote down the time.
By the end of this guide you will be able to explain why quiet months look idle. You will be able to name what the business can and cannot see. You will be able to recognise the trap of the heroic fix. You will be able to say what monitoring data can prove that memory cannot. And you will be able to keep a one-line-per-event work log you can copy today and read back a year from now.
In this guide
- The asymmetry
- What the business sees, and what it cannot
- Why "no news is good news" does not survive a budget review
- The heroic response trap
- What monitoring data can show that memory cannot
- The three habits
- The work log
- A quiet year at Meridian Parts
- The checklist
- Where to go next
The asymmetry
An outage is an event; its absence is not. That one sentence is the whole problem, and everything else in this guide follows from it. An outage is a period when a service people depend on is not working. It has a start time, an end time, a number of affected people, and usually a name that gets shouted across the office. The month in which the same service ran without interruption has none of those things. It has no start, no end, no headcount, and no name.
Events get remembered and non-events do not, because that is how attention works. Nobody notices the bridge that did not fall down; you notice the day it was closed. Your organisation is not being unfair when it remembers your one bad Thursday and forgets your fifty good ones. It is doing exactly what the shape of the information allows.
The asymmetry gets worse the better you are at the job. A team that catches problems late has plenty of visible events, most of them unflattering. A team that catches problems early produces fewer and fewer events until, from the outside, it produces nothing at all. The reward for excellent preventive work is a calendar with nothing on it. (Adequate preventive work gets you a calendar full of meetings, but at least people know where you sit.)
What the business sees, and what it cannot
The business sees three things about IT: tickets, outages, and the invoice. Tickets are the requests people log — the new starter, the broken printer, the forgotten password. They are visible because a person outside IT raised each one. Outages are visible because they hurt. The invoice is visible because someone signed it. None of the three measures whether the infrastructure is being looked after.
What the business cannot see is the work that happened because the monitoring system said something. It cannot see the alert acknowledged at six in the morning, from the kitchen, before anyone noticed that the mail server's queue was growing. It cannot see the disk that was expanded on a Tuesday because the graph said it would fill on Friday. It cannot see the switch port that had been flapping — going up and down repeatedly, usually a bad cable or a failing transceiver. That port was fixed before the users on it noticed anything worse than one slow morning.
Each of those is a real piece of work, and each one, by succeeding, erased the only thing that would have made it visible. The alert acknowledged at six prevented the outage at nine. The outage at nine would have been a ticket, a meeting, and a line in the monthly report. The acknowledgement is a timestamp in an alert history table that nobody outside IT has ever opened. This is not a problem you can solve by working harder, because the harder you work, the fewer events there are to see. It is a problem of record, and it is solved by keeping one.
Why "no news is good news" does not survive a budget review
"No news is good news" is true right up until someone has to decide what to spend next year. On an ordinary day it is a fair summary. It stops being true the moment a budget review asks what the monitoring licence, the second administrator, or the spare switch is for. That is because the honest answer — "so that you keep not hearing from us" — is an answer without evidence.
A budget review is a comparison of cost against value, and cost is always visible. The licence has a price. The salary has a number. The value side has to be supplied by you, and if you supply "nothing has broken", you have handed the review a blank. A blank is filled in by whoever is looking for savings, and to that person "nothing has broken" is indistinguishable from "nothing has happened".
The reviewer is not being unreasonable. Every other department has filled in its column with numbers: orders shipped, calls answered, invoices raised. IT's column says "it was quiet". The reviewer cannot approve a blank, and "quiet" is a blank wearing a tie.
The good news is that the value side is not actually blank. It is sitting in the monitoring database, timestamped. It includes every alert that was dealt with, and every graph that bent towards a limit and was bent back. It also includes every maintenance window that came and went without a complaint. It has simply not been written down anywhere the reviewer can read. Justifying monitoring spend covers the budget argument itself; this pillar is about the record that argument needs.
The heroic response trap
The heroic response is the outage fixed fast and visibly, and it is the most reliably rewarded thing an administrator can do. The service goes down at ten. You are on it by three minutes past. You find the failed disk, swap it, rebuild the array, and by half past eleven the service is back. People stop by your desk to say thank you. For a day you are the most visible person in the building.
The trap is that the same disk had been reporting rising errors for three weeks before it failed. And the administrator who replaced it in week one would have prevented the outage entirely and received nothing. From the outside, the heroic fix and the quiet prevention look like an eventful morning and an uneventful one. The eventful one gets the credit.
I have been on both sides of this, and the incentive is real. I am not suggesting anyone lets small things become visible things on purpose. I am saying that an organisation that only sees events will, over time, get more events, because that is what it is paying attention to. The cure is to make prevention as visible as repair, which is the job of this pillar.
A useful test for any piece of work: if this had gone the other way, what would the ticket have said? The disk at ninety-four per cent and rising becomes "file server full, nobody can save". The certificate expiring on Sunday becomes "website shows a security warning to every customer". If you can write the ticket that did not happen, you have found invisible work worth recording. Counting the outages you prevented sets the rules for when it counts, because the test is easy to pass generously.
If you are the whole department, the trap has a second jaw. In that case, you are the only witness to the quiet work and the only person who can be blamed for the loud kind. Solo admin burnout is about what that does to a person. The record is part of the remedy, because it moves the evidence out of your head.
What monitoring data can show that memory cannot
Memory is a poor witness for quiet work, and the monitoring system is a good one. Ask any administrator in December what they did in March and you will get a shrug and a guess. Ask the alert history and you get a list, with times. It was there, it was watching, and it wrote things down while you were too busy to. Four things in that data are evidence in a way that recollection is not.
Timestamps
Every alert has the time it was raised, the time someone acknowledged it, and the time it cleared. An acknowledgement is the click that says "a human has seen this and is dealing with it". And its timestamp is a record of a person at work. Six in the morning on a Saturday is a fact, not a claim.
Counts
How many alerts were raised this month, how many were acknowledged, and how many cleared on their own. Counts turn "we were busy" into "we handled forty-one", and they can be produced for every month in the same shape. A count on its own is thin; a count beside last month's count is a trend.
Graphs
A disk usage graph that climbs towards a hundred per cent and then drops on the day you expanded the volume is a picture of a prevented outage. So is the interface error counter that rises for a week and flattens the day you replaced the cable. The graph shows the problem, the intervention, and the outcome in one image. And a manager can read it without knowing what a volume is. Most of those catches come out of the weekly trend review, which is where you look at the graphs on purpose.
The before and after
A change made safely leaves a signature. The maintenance window opened. The device went quiet. The device came back. The checks went green. And no alert followed. That sequence, pulled from the history, is proof that the change was done well. Nobody remembers a change that went well. The monitoring system does, because it was just recording.
None of this is interpretation. It is data that already exists in your installation, and pulling it out is a query.
The three habits
Three habits make invisible work visible. They are habits rather than projects because a project ends and the invisibility does not.
Record it as it happens
Write down each piece of quiet work on the day you do it, in one line, with a pointer to the evidence. Not a report, not a paragraph: one line. Quiet work is forgotten within a week, by you as much as by anyone. A line written on the day costs a minute while a line reconstructed in December costs an afternoon and is wrong.
Summarise it monthly
Once a month, turn the lines into a short page: what was caught early, what was handled, what was changed safely, what is coming. This is the quiet month report, the IT-team version of the record. The monthly one-pager, with its four boxes, is the finance-facing cousin, and the quiet month report is what you build it from.
Tie it to a business outcome
Every entry should say what would have stopped. Not "disk expanded" but "disk expanded; invoicing would have stopped saving on Friday". The reader does not know what the file server does, but they know what invoicing is. Speaking business, not packets teaches the translation; the habit is to do it while writing the line, when you still know the answer.
Skip the first habit and the other two have nothing to work from. Skip the second and the log is a private diary. Skip the third and the record is accurate, complete, and unread.
The work log
The work log is a table with one line per event, kept for a rolling year. It records what the monitoring showed, what you did about it, what would have happened otherwise, and where the evidence is. It is the raw material for everything else in this pillar. It does not replace a ticket system; tickets record what people asked for, and the work log records what the monitoring system asked for.
| Date | What monitoring showed | What you did | What would have happened | Evidence |
|---|---|---|---|---|
| (date) | File server data volume at 94 per cent, rising about 1 per cent a day | Extended volume by 500 GB in the Tuesday maintenance window | Volume full within a week; shared drive and invoicing exports stop saving | Disk usage graph, 30 days; change record |
| (date) | Mail queue length alert at 06:10, queue growing | Acknowledged 06:14; restarted stuck transport service; queue drained by 06:40 | Outbound mail delayed for the whole office from 08:00 | Alert history entry; queue length graph |
| (date) | Interface errors rising on access switch port 14 over five days | Replaced patch lead; errors flat since | Port flap taking the finance printer and three desks offline intermittently | Interface error graph, 14 days |
| (date) | Certificate on the customer portal expiring in 14 days | Renewed and installed; check green | Browser warnings on the public site from the expiry date | Certificate check history |
| (date) | Backup job overran and did not complete | Found the backup target full; cleared old sets; re-ran and verified | No usable backup of the finance database until someone noticed | Backup check history; job log |
Five rules keep the log honest and cheap to maintain.
- One line, on the day. If the line takes more than two minutes to write, you are writing the report, not the log.
- The evidence column is a pointer, not the evidence. Name the graph, the alert entry, or the change record. The monitoring system is keeping the graph for you.
- "What would have happened" is specific and modest. Name the service and the people, and stop. "Invoicing exports stop saving" is a claim you can defend. "Business-critical impact" is not a claim at all.
- Keep a rolling year. Anything older goes into an archive sheet, so the annual review, whenever it falls, can see twelve months in one place.
- Keep it where a colleague could find it. A spreadsheet in the team's shared documentation, not a file on your desktop and not, under any circumstances, your memory.
The monthly counts to go alongside the log come from the alert history. The shape of the query is the same for most monitoring databases even though table and column names differ. So treat this as illustrative and check the product documentation for yours:
-- Illustrative: adapt table and column names to your monitoring database
SELECT strftime('%Y-%m', raised_at) AS month,
COUNT(*) AS raised,
SUM(CASE WHEN acknowledged_at IS NOT NULL THEN 1 ELSE 0 END) AS acknowledged,
SUM(CASE WHEN acknowledged_at IS NULL
AND cleared_at IS NOT NULL THEN 1 ELSE 0 END) AS cleared_on_their_own
FROM alert_history
WHERE raised_at >= DATE('now', '-12 months')
GROUP BY month
ORDER BY month;
The three numbers mean different things. "Raised" is the headline count, and much of it is noise. "Acknowledged" is the number of times a person looked at something and did something, which is the number that measures work. "Cleared on their own" says something about your thresholds rather than your effort, and an alert that cleared itself is never counted as a save.
Tip: keep the work log next to the uptime evidence file. Building the uptime evidence file describes the availability side of the same idea. The two documents answer the two halves of one question: what stayed running, and what you did to keep it that way.
A quiet year at Meridian Parts
At Meridian Parts, a single administrator ran the network for a year in which nothing went down. She achieved that by expanding two volumes and replacing a failing core switch power supply on a Saturday. She also achieved it by renewing four certificates and acknowledging somewhere over a hundred alerts before anyone else was awake. She kept none of it written down, because she was busy, and because it had not occurred to her that it needed writing down. The year-end summary for the board described the period as "nothing much happened in IT", and the following year's budget was set accordingly. She now keeps a work log. The summary the year after read "IT prevented nine service interruptions", which is the same kind of year, described by someone who had the list.
The board was not careless. It summarised what it could see, and what it could see was nothing. The work log is not a way of taking credit; it is a way of leaving a trace that the work was done. And it has to exist before the question is asked, because it cannot be built in the meeting.
The checklist
Use this once a month, on the day you write the summary.
- A work log with the five columns exists, and it has at least one entry from the last two weeks.
- Every entry was written on the day, in one line.
- Every entry says what would have happened, naming a service and the people who use it.
- Every entry points to evidence that exists in the monitoring system or the change record.
- You can produce alert counts per month for the last twelve months, separating acknowledged from cleared-on-their-own.
- A monthly summary date is on the calendar.
- You can state one prevented outage from the last quarter, with its evidence, in a sentence a non-technical reader would understand.
- The log is stored where a colleague could find it if you were away.
- Entries older than a year have been moved to the archive sheet rather than deleted.
Where to go next
Quiet months look idle because an outage is an event and its absence is not. The monitoring system has the timestamps, the counts, the graphs, and the before-and-after of every quiet save you made this year. Three habits turn that into a record: write the line on the day, summarise it monthly, and say what would have stopped. A year of those lines is the difference between "nothing much happened" and "nine interruptions prevented".
The next guides build on the log. The quiet month report is the monthly page. Uptime reporting as reputation defence is the standing availability record. Counting the outages you prevented sets the rules for the strongest claim you can make. Answering "what do we pay you for?" is the page that ends the conversation. All are part of Making Invisible Work Visible With Monitoring Data.
Put this into practice with SysPeep Monitor
A work log is only as good as the history behind it. SysPeep Monitor keeps alert history, availability reports, and scheduled reports in one web console. It is fed by a central Windows service polling over SNMP, WMI, and SSH. So the timestamp for the alert you acknowledged at six in the morning is still there when you write the line. The free trial installs on a Windows machine if you want to see what a month of your own quiet work looks like.
Download Free TrialLearn About SysPeep