Writing the Monitoring Budget Line: Licence, Hardware, Time, and Training
Ask most IT teams what monitoring costs and they will tell you the licence. Ask finance the same question after a bad year and they will tell you something considerably larger. They will have added the server, the storage, and the fortnight someone spent rebuilding thresholds. They will also have added the standby allowance for whoever carries the phone. Both answers are honest. Only one survives the meeting.
The budget line most teams write is the smallest part of the cost. And a line that understates itself is a line that gets cut. The moment someone discovers the rest of the cost, the whole line looks like it was hiding something. I learned this by budgeting the licence and forgetting the server. It had been "free" because it was old, right up until it was old enough to die during the renewal week.
By the end of this guide you will be able to state the full annual cost of your monitoring honestly, in five parts. You will be able to measure the one part nobody measures, which is people's time. You will be able to extend the line to three years so the renewal surprises nobody. You will know what can be trimmed and what must not be. And you will be able to present the line next to the dividend and one story so it is not the first thing cut. There is a table template and a one-paragraph justification to fill in.
In this guide
- The licence is the smallest part
- The full annual cost, in five parts
- Measuring people's time
- The three-year view
- What can be trimmed, and what never should be
- Presenting the line so it is not the first cut
- The budget line template
- A war story
- The budget line checklist
- Where to go next
The licence is the smallest part
A monitoring budget line has five parts, and the licence is usually the smallest once the others are counted. That is not an argument against the licence. It is an argument for writing down the other four. A leader who sees only the licence will assume that is the cost, and will be annoyed twice. The first time is when the server needs replacing. The second is when someone mentions that the administrator spends a day a fortnight on the thing.
The purpose of stating the full cost is not to make the line bigger. It is to make it true. When you put the dividend from the downtime math guide next to it, the comparison is between real benefit and real cost. Then nobody can knock it over by finding a cost you left out. A true line that the dividend exceeds is unassailable. A small line that the dividend exceeds by more is waiting for someone to find the server.
This guide is about the annual line for monitoring you already run. Comparing the whole-life cost of one product against another is a different exercise, and the guide on total cost of ownership covers it. The two count similar parts; they differ in that this one has to be defended every year.
The full annual cost, in five parts
Count all five, every year, even the ones that come to nothing. A zero you wrote down is a decision and a blank is an oversight.
1. The licence or subscription
This is the figure on the renewal quote, and it is the easy one. Note the basis it is charged on: per device, per sensor, per probe, or a tier with a ceiling. The basis matters more for next year than this one, because your device count will change and the line has to change with it. The guide to licensing models explains what each basis means for a growing network. For the budget line, write the basis in the notes column so the person reading it next year knows why the number moved.
2. The server it runs on, and its share of everything
The monitoring server is a real machine, or a real slice of a virtual host. It consumes storage, backup capacity, power, and a maintenance window like any other. Its share of those is a cost, even if no invoice ever says "monitoring" on it. Count it as a fraction. That is the server's share of the host it lives on, of the storage it uses, and of the backup system that copies its database every night. If the database backend is a separate server, or a share of one, count that too. A fair approximation is fine; a zero is not.
The storage share is easy to measure. On the monitoring server, this reports the size of the data folder and its backups in gigabytes. Multiply that by whatever your organisation reckons a gigabyte of backed-up storage costs a year:
# Size of the monitoring data folder and its backup copies, in GB,
# for the storage-share line. Adjust the paths for your install.
Get-ChildItem -Path "D:\Monitoring\Data", "E:\Backups\Monitoring" -Recurse -File |
Measure-Object -Property Length -Sum |
ForEach-Object { [math]::Round($_.Sum / 1GB, 1) }
3. People's time
This is the part nobody counts and the part that is usually largest. Monitoring needs upkeep: adding new devices, retiring old ones, reviewing thresholds, pruning alerts, tending dashboards, producing the monthly report, applying the product's updates. None of it is dramatic and all of it is hours. The next section is about measuring it, because a guess is the first thing a finance director will challenge. The guide on the hidden costs of cheap monitoring is honest about how this time balloons when the tooling is fragile. The point here is only to count it.
4. Training and the lab
Someone has to know how the product works, and knowledge costs either a course or the hours spent learning by doing. Count both: the training itself, if any, and a realistic number of hours per year. Those hours cover keeping up with the product and onboarding anyone new. A lab is part of this. A small install lets you try a threshold change or an upgrade before it touches production. It needs a machine and, for some products, a licence. A product with a free trial can serve as the lab for a while, and you can say so on the line.
5. The cost of the alerts themselves
Alerts go to people, and people out of hours are a cost. On-call is the arrangement where a named person carries the phone outside working hours and responds to alerts. If your organisation pays a standby allowance or grants time off in lieu for it, that is part of the monitoring line. Without monitoring there would be nothing to carry the phone for. If nothing is paid, the honest line notes the out-of-hours hours anyway, from the alert log. That way, the ask is truthful about what it costs the people as well as the budget. The guide on humane on-call is about keeping that number bearable.
Measuring people's time
Measure the time for a fortnight and multiply. A fortnight is long enough to catch the weekly rituals and short enough that people will actually keep the log. Everyone who touches the monitoring keeps one line per task for ten working days: date, who, category, minutes. The categories are fixed so the totals are comparable next year.
Monitoring time log (two working weeks, everyone who touches it) Date | Who | Category | Minutes | What -----------+-------+-------------+---------+----------------------------- [date] | [ini] | upkeep | 45 | added 3 switches, retired 1 AP [date] | [ini] | thresholds | 30 | reviewed disk alerts, raised 2 [date] | [ini] | alerts | 20 | tuned flapping interface rule [date] | [ini] | dashboards | 15 | fixed dead panel on wall screen [date] | [ini] | reports | 60 | monthly one-pager [date] | [ini] | product | 90 | applied update, checked probes [date] | [ini] | training | 40 | walked new starter through maps Annual hours = (total minutes / 60) * 26 Annual cost = annual hours * loaded hourly cost (from finance)
Multiply by twenty-six because the log covers two weeks. Do not adjust for holidays or "a quiet fortnight". If it was quiet, say so in the notes and run it again in a busy one. The loaded hourly cost is the figure your finance team uses for a person's full cost per hour. They will supply it if asked. Do not construct your own. Two fortnights a year, averaged, is a defensible number. One fortnight, honestly labelled, is still better than an estimate, and an estimate is better than the blank most teams submit.
Expect the number to be larger than you thought. That is normal, and it is not an argument for cutting the monitoring. It is the true cost of the thing that produced the dividend. (It is also the first time anyone above you will have seen that the monitoring does not run itself.)
The three-year view
A budget line that shows only this year hides the three things that make next year's line jump: renewal growth, the hardware refresh, and headcount change. Show three years, with the second and third marked as projections, and the jumps become plans instead of surprises.
Renewal growth. Your device count grows, and the licence basis turns that growth into cost, sometimes smoothly and sometimes in a step when you cross a tier. The guide on licensing growth traps lists the ways a renewal jumps. For the line, project the device count from your own inventory growth over the last year and write the resulting tier in the notes. Negotiating renewals covers what to do about the quote when it arrives. Meanwhile, the hidden-costs checklist catches the add-ons that appear on it. A renewal quote is a document that has never once gone down of its own accord.
Hardware refresh. The monitoring server has a lifetime like any other, and the year it is replaced is a year with a larger line. Put the refresh in the year it is due, not the year it fails. The second option is the same money plus an outage of the thing that tells you about outages.
Headcount change. A second administrator means training and lab time again. Losing the one who built the monitoring means someone else learning it, which is training time plus the mistakes. Neither is predictable to the year. But a line that says "training: assumes no staff change" has thought about it. The note is what makes the number credible when it moves.
What can be trimmed, and what never should be
When the line has to come down, three things can be trimmed with limited harm, and two must not be touched at any saving. Knowing the difference in advance lets you say "yes, we can reduce it, here is how" in the meeting. The alternative is "no", which is the answer that gets the line cut for you.
Trim sensor counts first. Every device has interfaces and metrics nobody looks at, and on a per-sensor basis each one is licence and polling load. A quarterly pass that removes the unused ones often takes the licence down a tier without losing anything anyone would notice. Trim retention second. Keeping every raw sample for years is storage share, and rolled-up history usually answers every question the raw data did. For that decision, the metric retention policy guide shows how. Trim redundant tools third, often the biggest saving. If two products watch the same switches, one of them is a line with no dividend. Use the tool inventory guide to find out which.
Do not trim the monitoring server's own disk and backups. Not to save a disk. Not "just until the refresh". A monitoring server with no backup is a year of evidence that vanishes with one failed drive. And the evidence is most of what the line is for. See the guide on backing up the monitoring system; it is short and worth the afternoon. And do not trim the person's review time. The hour a month spent reviewing thresholds and pruning alerts is what keeps the alerts true. Monitoring with untrue alerts is a line with a dividend of zero and an on-call cost that keeps rising. The threshold review ritual is that hour, described.
Tip: prepare the trimmed version of the line before the meeting, with what each trim loses written beside it. A leader who is offered "the full line, or this smaller one that gives up raw history older than a quarter" is choosing between two things you designed. A leader who is offered only the full line is choosing between it and nothing.
Presenting the line so it is not the first cut
A line on its own is a cost, and costs get cut. A line next to the dividend it produced and one story about what that looked like is an investment with a return. And investments with a return get argued about, which is a much better fate. Never submit the line without the other two on the same page.
The layout is three things in order. The full annual cost, in five parts, with the three-year view beneath it. The dividend, shown with and without the soft row, and the sentence that says which assumption would break it. Then one four-sentence story from the before-and-after guide, with its ticket number and alert timestamp. Cost, return, proof. If there is room for one more thing, it is a single comparison with a similar organisation, and only one. See the guide on benchmarks in budget conversations for why more than one starts an argument you do not need.
Say the trimmed version exists, and what it gives up, before anyone asks for cuts. It shows the line was built by someone who has already looked for savings. It moves the conversation from "can this be smaller?" to "which of these two do you want?". Lines that arrive with their own trimmed alternative are rarely the first cut, because cutting them requires reading them. The first cut is usually the line nobody had to read.
The budget line template
One row per part, with the basis written down so the number can be recalculated next year by someone who is not you. Fill every cell; write "0, see note" rather than leaving a blank.
| Item | Basis | This year | Next year | Notes |
|---|---|---|---|---|
| Licence or subscription | Renewal quote; licensed on [devices / sensors / probes / tier] | Current count, tier ceiling, projected count at renewal | ||
| Monitoring server | Share of host or VM, storage (GB measured), backup, power | Refresh due [year]; database backend included or separate | ||
| Remote probes | Count, plus any per-probe licence and the host each runs on | Sites covered | ||
| Staff time: upkeep and product updates | Fortnight log, category totals, times 26, times loaded rate | Log dates; quiet or busy fortnight | ||
| Staff time: reviews and reports | Fortnight log as above | Review hour not trimmable; see justification | ||
| Training and lab | Course or hours; lab machine and licence | Assumes [no / one] staff change | ||
| On-call | Standby allowance or time in lieu; out-of-hours alert hours from the log | Alerts out of hours last year: [n] | ||
| Total | Sum of the above | Three-year view attached | ||
| Dividend | From the downtime math worksheet | With and without avoided incidents | ||
| Trimmed alternative | Total after the listed trims | What each trim gives up |
Beneath the table goes one paragraph, written for the person who will read the line without you in the room. It contains the cost, the return, the proof, and the ask, and nothing else:
The monitoring line for [year] is [total], of which the licence is [amount] and staff time is [amount]. This year monitoring found [n] of [m] faults before users reported them, caught [k] capacity limits before failure, and supplied the evidence for [j] provider or audit questions. The dividend, calculated from our own incident log using the cost-per-minute figures supplied by finance, is [amount], and exceeds the line under [which sensitivity rows]. [Four-sentence story, ticket and alert timestamp attached.] We are asking for [total] to keep this in place. A trimmed line of [amount] is possible by [trims], at the cost of [what is lost]. Without the line, faults will be found by users, and the evidence file stops growing on [date the licence lapses].
A war story
Pennant Foods cut the monitoring licence in a lean year, on the reasoning that the server would keep running and the licence was "mostly support". It did keep running, in the sense that the service started. It stopped collecting when the licence lapsed, and the dashboards showed the last good day for five months. On a holiday weekend the goods-in scanner network lost its uplink on the Friday evening and was found on the Tuesday, when the first lorry arrived. Receiving was done on paper for two days. A chilled delivery was refused because it could not be booked in, and an agency was paid to re-key the backlog. Finance costed the incident at roughly three times the annual licence, using their own figures. The line was reinstated at the next round with a note describing it as "reviewed".
The budget line checklist
Run the line through this before it goes to anyone. Any no is a question you will be asked.
- All five parts appear: licence, server share, people's time, training and lab, on-call.
- The licence basis is written in the notes, with the current count and the tier ceiling.
- The server share includes measured storage and backup.
- People's time comes from a fortnight log, with the dates and the loaded rate's source.
- The on-call row shows either the allowance or the out-of-hours alert hours from the log.
- The three-year view shows the renewal projection, the refresh year, and the headcount assumption.
- The dividend sits on the same page, shown with and without the soft row.
- One before-and-after story is attached, with its evidence.
- A trimmed alternative is prepared, with what each trim gives up.
- The monitoring server's backups and the review time are marked as not trimmable, with the reason.
- The justification paragraph contains the cost, the return, the proof, and the ask, and nothing else.
Where to go next
The licence is the smallest part of the monitoring line. The line is only defensible when all five parts are on it: licence, server share, measured time, training, and the cost of the alerts themselves. Show three years so the jumps are plans, and prepare the trimmed version yourself. Protect the backups and the review hour. Never let the line travel without the dividend and one story beside it. A true line with a return attached is rarely the first cut, because cutting it means arguing with arithmetic.
What monitoring spend actually buys supplies the plain sentences for the justification paragraph. Meanwhile, the downtime math guide supplies the dividend row. And when leadership says no is for the year the line is cut anyway, and what to keep running when it is. All of them are part of Justifying Monitoring Spend to Leadership.
Put this into practice with SysPeep Monitor
The server-share and time rows of the line are easier to keep small when the monitoring itself is one central Windows service with optional remote probes. It is backed by SQLite, MySQL or MariaDB, or SQL Server. It is administered from a single web console for dashboards, reports, and alert history. That is how SysPeep Monitor is built. The free trial at /download/ can serve as the lab install on the training row while you decide.
Download Free TrialLearn About SysPeep