Benchmarking Monitoring Spend and Staffing Without Inventing Numbers
Somebody will eventually ask whether you spend too much on monitoring. Or they will ask whether two administrators is the right number for a network your size. They will ask it in a tone that suggests they already have a figure in mind. They usually do. It came from a slide and has no definition attached. And it is about to be compared with a number you have never actually calculated. This guide is about calculating it first, honestly, so that the conversation starts from your figures rather than theirs.
You will learn four ratios that can be computed from data you already have. The first three are devices per administrator, monitoring spend as a share of infrastructure spend, and alerts per person per week. The fourth is hours per week spent keeping the monitoring running. You will learn how to count each one without flattering yourself, including the costs that hide in people's time. There is a fully worked example with fictional inputs and a worksheet to keep. There is also a list of the questions to ask before you conclude anything from a ratio that looks wrong.
There is no table of typical values in this guide. The ratios are for comparing you with yourself over time and, carefully, with a peer whose definitions match. They are not for comparing you with an industry that has never been measured the same way twice.
In this guide
- The four ratios worth computing
- Computing each one honestly
- A worked example: Northgate Retail
- The spend and staffing worksheet
- What a comparison can support, and what it cannot
- When the ratio looks wrong: questions before conclusions
- A war story and a checklist
- Where to go next
The four ratios worth computing
A ratio is one of your numbers divided by another of your numbers. And the reason to use ratios rather than raw figures is that ratios survive growth. A licence bill that doubled is alarming; a licence bill that doubled while the device count tripled is a bargain. Four ratios cover nearly every spend and staffing question a manager will ask. All four can be built from your monitoring tool, your renewal paperwork, and a fortnight of honest notes.
Monitored devices per administrator is the number of devices your monitoring actually watches, divided by the number of people responsible for it. Adjust that divisor for how much of their time that responsibility takes. Monitoring spend as a share of infrastructure spend is everything monitoring costs in a year, divided by everything the infrastructure costs in the same year. Alerts per person per week is the count of alert notifications sent in a week, divided by the people who received them. Hours per week on monitoring upkeep is the time spent keeping the monitoring itself healthy: adding devices, tuning thresholds, fixing the collector, reading the reports. Not the time spent fixing what the monitoring found; that is operations, and mixing the two is the first way this measure goes wrong.
Two of the four are about people, and that is where the surprises live. Licence figures are on an invoice. Time is remembered, and memory is generous to whichever story the rememberer prefers.
Computing each one honestly
Devices per administrator
Take the device count from the monitoring tool's own discovery inventory. Do not take it from the licence tier or from the spreadsheet somebody made when the network was installed. The inventory is the only count that a peer can produce the same way. And it is usually larger than anyone expected. It includes the printers, the access points, the desk phones, and the three switches nobody remembers buying. Then decide what an "administrator" is. If two people share monitoring with everything else they do, the honest divisor is not two. It is the fraction of their time the monitoring takes, added up. A quarter of one person and a quarter of another is half a person. The ratio is devices per half-person, which is a very different number from devices per two.
The fraction comes from the fortnight of counting described under the fourth ratio below, so compute that one first. Dividing by heads when you should divide by time is the most common error in staffing comparisons. It flatters a team that does many things and punishes one that does monitoring well, which is precisely backwards.
Monitoring spend as a share of infrastructure spend
Monitoring spend has three parts, and most people count one. The first is the licence or subscription, which is on the renewal paperwork. You may not be sure what you are actually paying for, or whether the count on the licence matches the count in the tool. In that case, Tracking Licenses and Usage sorts that out first. The second is the hardware or virtual machine the monitoring runs on, plus its share of storage and backup. It is a modest figure, but not zero, and not nothing when the database has grown for three years. The third is people's time, at whatever cost your finance team uses for an hour of an administrator. Multiply that cost by the hours per week from the fourth ratio and by the weeks in a year. This third part is regularly the largest, and it is the one the slide never includes. Hidden Costs of Cheap Monitoring is the longer argument for why it belongs in the total.
Infrastructure spend is the denominator, and it needs the same honesty in the other direction. It includes servers, storage, network hardware, the internet lines, the supporting software, and the people who run it all, for the same year. Use whatever figure your finance team already reports for infrastructure, because a ratio computed with the official denominator is one nobody can argue with. Then divide. The result is a share, and the share on its own tells you nothing. The share compared with last year's share, computed the same way, tells you whether monitoring is growing faster or slower than the estate it watches.
Alerts per person per week
Count from the alert history, over a whole quarter, and take the median week rather than the average. That way, one dreadful weekend does not define the period. Count every channel that reaches a person: email, text messages, whatever the phone app does. Do not count alerts that went to a mailbox nobody reads, but do write down that such a mailbox exists. It is one of the signs in Healthy Signs. Divide the weekly count by the people on the notification list. A rising trend in the result is the earliest warning that the alerting needs attention.
Hours per week on monitoring upkeep
Count for a fortnight. Not a day, because a day is not representative, and not a quarter, because nobody keeps a time log for a quarter. Every person who touches the monitoring writes down, at the end of each day, the time they spent on it. That includes adding or removing devices, adjusting thresholds and alert rules, and fixing the collector or the database. It also includes building or reading reports and answering "is this alert real?" from colleagues. Anything spent fixing the thing the alert was about goes in a different column. Two weeks of this is mildly annoying and completely persuasive. A spreadsheet with a date, a person, a category, and a duration in hours is all it needs. The weekly average is a single formula:
Fortnight log columns: A=Date B=Person C=Category D=Hours Weekly hours on monitoring upkeep: =SUMIF(C:C,"monitoring",D:D)/2 Fraction of one person's time (37.5-hour week as the example working week): =SUMIFS(D:D,B:B,"Alice",C:C,"monitoring")/2/37.5
The second formula gives the fraction you need for the devices-per-administrator divisor. Substitute your own working week; the number I have used is an example, not a recommendation. I have kept this log at three organisations. Each time the people who thought they spent "hardly any time" on the monitoring turned out to spend an afternoon a week. And the person who thought it ate their life was spending about the same. Memory is not a measuring instrument.
Tip: tell the team the fortnight log is for a ratio, not for judging anyone, and mean it. A time log that people fear becomes a work of fiction by the third day. And you will have measured nothing except how quickly your colleagues can learn a new form.
A worked example: Northgate Retail
Northgate Retail is fictional, and every figure below is an invented input chosen to show the arithmetic. It has a head office and a handful of shops, and the discovery inventory in its monitoring tool lists 420 monitored devices. Two administrators, Alice and Ben, look after everything. The fortnight log shows Alice spent 12 hours on monitoring upkeep across the two weeks and Ben spent 10. The alert history for the last quarter has a median week of 96 notifications, all sent to both. The licence renewal is some figure; call it L, because the currency and the amount do not matter to the method. The virtual machine, storage, and backup share is a smaller figure; call it H.
| Ratio (Northgate Retail, fictional inputs) | Calculation | Result |
|---|---|---|
| Weekly upkeep hours | (12 + 10) / 2 | 11 hours per week |
| Monitoring time as fraction of a person (37.5-hour week) | 11 / 37.5 | about 0.29 of one person |
| Devices per administrator (by heads) | 420 / 2 | 210 |
| Devices per monitoring full-time equivalent | 420 / 0.29 | about 1,450 |
| Alerts per person per week (median) | 96 / 2 | 48 |
| Annual people cost of upkeep | 11 × 52 × hourly cost | 572 hours × hourly cost |
| Monitoring spend, total | L + H + (572 × hourly cost) | your figures |
| Cost per monitored device per year | total / 420 | your licence total plus the rest, divided by the device count |
| Share of infrastructure spend | total / infrastructure total | a share; compare with last year's share |
Two things in that table are worth a second look. The devices-per-administrator figure changes by a factor of about seven depending on whether you divide by heads or by time. Only one of those figures describes the monitoring workload. And the annual people cost, 572 hours in this example, is a number that rarely gets written down anywhere. That is why the licence looks like the whole cost of monitoring until somebody does the fortnight log. Suppose Northgate's total, people included, works out to one twentieth of its infrastructure figure. In that case, the share is five percent, and it means nothing on its own. If last year's share, computed the same way, was four percent, monitoring is growing faster than the estate. That would be a sentence for the report. If it was six, it is a different sentence.
A full-time equivalent is simply "one person's whole working week"; two people at half time are one. It is the unit that makes staffing ratios comparable. You can compare a team where monitoring is a job with a team where it is a fraction of several jobs.
The spend and staffing worksheet
The worksheet below is the version of the table that you keep. It has one row per ratio and a column for where each input came from. A ratio whose inputs cannot be traced will be challenged and deserves to be. This period and last period sit side by side. There is also a column for the caveats you would want a stranger to read before they quoted the number back at you.
| Ratio | Formula | Inputs and source | Last period | This period | Direction | Caveats |
|---|---|---|---|---|---|---|
| Devices per monitoring full-time equivalent | monitored devices / summed time fraction | discovery inventory; fortnight log | [value] | [value] | [up / flat / down] | [what is in the device count; what "administrator" includes] |
| Monitoring spend as share of infrastructure spend | (licence + hardware share + people hours × hourly cost) / infrastructure spend | renewal; finance figure for infrastructure; fortnight log | [share] | [share] | [up / flat / down] | [what the finance denominator includes] |
| Cost per monitored device per year | monitoring spend / monitored devices | as above; discovery inventory | [value] | [value] | [up / flat / down] | [device mix; licence tier boundaries] |
| Alerts per person per week (quarter median) | weekly notifications / recipients | alert history export | [value] | [value] | [up / flat / down] | [channels counted; any unread mailbox] |
| Hours per week on monitoring upkeep | fortnight upkeep hours / 2 | fortnight log | [value] | [value] | [up / flat / down] | [what counted as upkeep; who logged] |
Recompute it at every renewal and whenever headcount changes, and file each version with the monthly report rather than overwriting it. A single ratio is a fact. Five renewals of the same ratio, computed the same way, is a trend, and trends are what budget conversations are actually about.
What a comparison can support, and what it cannot
A trend in your own ratios supports a claim about direction, and that is the strongest claim these figures can make. "Cost per monitored device has fallen at each of the last three renewals while the device count grew" is a sentence built entirely from your own data under one method. Nobody in the room can produce a slide that contradicts it. "Hours per week on upkeep doubled after the last site was added" is the same kind of sentence, pointing the other way. It is the start of an honest conversation about automation or staffing.
A comparison with a peer whose definitions match supports a conversation, which is a smaller thing than a conclusion. If a peer of similar size and hours has a much lower upkeep figure, the useful output is a phone call. In that case, ask what they automated, rather than making a decision. If their alerts per person are a third of yours, the question is what their thresholds look like. The comparison does not tell you that you are wrong. It tells you where to look.
What none of these ratios can support on its own is a headcount case. "Our devices per administrator is higher than our peers, therefore we need another person" skips every step that matters. Those steps are what the extra person would do, what happens if they are not hired, and what the alternatives cost. Those steps are a budget case, and Building a Budget Case From Monitoring Data shows how to build one. The ratio is evidence that the request is reasonable. The case is the request. Confusing the two is how a good administrator ends up defending a number instead of a need.
When the ratio looks wrong: questions before conclusions
A ratio far from a peer's, or from where it was last year, is a prompt for questions, and the questions come before any conclusion. These are the ones I ask, in order.
- What is in the device count? A site with hundreds of desk phones and printers has a flattering devices-per-administrator figure and a workload that has nothing to do with it. Compare the mix, not just the count.
- What hours are covered? A team that carries a phone at night is doing a different job from one that does not. And their hours-per-week figure includes work the other team's never will. Match the hours before comparing the people.
- How much is automated? Auto-discovery, templated checks, and dependency rules reduce upkeep hours without reducing the device count. A peer with a lower upkeep figure may have configured their way there, which is a lesson, not a reproach.
- How many sites, and how far apart? Remote sites cost travel, probes, and patience, and none of that shows in a device count.
- What does "administrator" include? If the person counted also runs the helpdesk, the fraction is wrong. Go back to the fortnight log.
- Did the definition change? A jump between periods is very often a change in what was counted. Check the caveats column before checking anything else.
If you are the only administrator, the ratios still work. In that case, the time-fraction divisor matters even more, because "one person" is never one full-time equivalent of monitoring. The Solo Admin's Monitoring Setup covers keeping upkeep small enough that the fraction stays honest. And if the ratios say spending is rising faster than the estate, Where Monitoring Spending Actually Helps answers where money changes outcomes. That is rarely "cut the licence and double the hours".
A war story: the night shift that was not on the slide
Kestrel Foods ran a cold-storage operation with two sites, three administrators, and a warehouse that never closed. A consultant's review concluded that Kestrel had "too many administrators per device" compared with a figure from somewhere unstated. The administrators, who had kept a fortnight log the year before, brought it to the meeting. It showed that the third administrator's hours were almost entirely nights and weekends, covering the freezer alarms and the conveyor network. The log also showed that during business hours the monitoring was looked after by roughly half of one person. Divided by time rather than heads, Kestrel's ratio was unremarkable, and the review's figure had come from organisations that switched their alerting off at six. The consultant revised the slide, and the freezers, which had never been consulted, carried on at their usual temperature.
Nobody in that story was dishonest. The consultant used the only figure available, and it had been built by dividing by heads. The fortnight log was the only measured evidence in the room, and it won without needing to be loud.
The checklist
Work through this before you compute a spend or staffing ratio, and again before you quote one.
- The device count comes from the monitoring tool's discovery inventory, and I know roughly what kinds of devices are in it.
- A fortnight time log has been kept by everyone who touches the monitoring, with upkeep separated from fixing what the monitoring found.
- Staffing ratios divide by the summed time fraction, not by heads.
- Monitoring spend includes the licence, the hardware and storage share, and the people hours at a cost figure finance agrees with.
- The infrastructure denominator is the figure finance already reports, for the same year.
- Alerts per person per week is a quarter's median from the alert history, across every channel that reaches a person.
- Every ratio has a "last period" beside it, computed the same way, and a caveats column that is not empty.
- Any comparison with a peer was preceded by an exchange of definitions.
- The ratio is presented as evidence that a request is reasonable, never as the request itself.
Where to go next
Spend and staffing is one of the three comparisons that Am I Normal? What Benchmarking Can and Cannot Tell You singles out. The uptime comparison, with its definitions-first email and its own worked example, is in Benchmarking Uptime and Incident Counts Fairly. For the parts of a practice that no ratio captures, Healthy Signs describes them in words. And when a ratio from this guide has to sit in a finance document next to a request, Using Benchmarks in Budget Conversations covers how to phrase it. All of them live under Am I Normal? Benchmarking Your Network and Monitoring Practice.
Put this into practice with SysPeep Monitor
Every ratio in this guide divides by a device count. The honest count is the one the monitoring discovered rather than the one on the licence. SysPeep Monitor auto-discovers devices and interfaces. It keeps alert history as a report in the same web console as availability. It runs as one Windows service with optional remote probes for the other sites. So the inputs to your worksheet come from one place. The free trial installs in minutes.
Download Free TrialLearn About SysPeep