Home › Topics › Building a Monitoring Strategy › Goals and Scope

Start With Why: Defining Monitoring Goals and Scope Before Touching a Tool


Walk up to a monitoring system that has been running for a few years and you are looking at archaeology. There is a layer of ping checks from the winter the core router died. Above that, a stratum of disk-space alerts laid down the week the mail server filled up. A scattering of certificate checks marks the Monday an expired certificate made the login page terrifying. Nobody designed any of it. Each layer was added in the days after an incident, by whoever was on duty, so that one exact embarrassment could never happen again. I have inherited two systems like this, and I built a third.


Every individual check in a system like that is reasonable. The problem is the shape of the whole. Coverage mirrors your incident history instead of your business, so you are superbly protected against the last five outages and blind to the next one. Nobody can answer the simple questions, either. What is this monitoring for? Is anything important missing? How would we know when we were done?


There is an alternative, and it costs a pad of paper. Before you install anything, evaluate anything, or add one more check, you answer three questions in writing. This guide walks through those questions and the one-page scope statement they produce. It covers how to interview the rest of the company so your answers are honest, and the classic mistakes to dodge. By the end you will have the founding document that everything else in your monitoring strategy builds on. It takes about twenty minutes to write. I put mine off for years, and every one of those years added a layer.


In this guide

  • How monitoring becomes archaeology
  • The three questions to answer in writing
  • The one-page scope statement
  • Why the exclusions are the powerful part
  • Stakeholder interviews for a small org
  • From goals to a check philosophy
  • Three anti-patterns to avoid
  • A worked example: 120 people, one page

How monitoring becomes archaeology


The loop that builds these accidental systems is completely natural. Something breaks. Someone asks the question every admin dreads — "why didn't we catch this?" — and the honest answer is that nothing was watching. So a check gets added, the incident is closed, and everyone moves on. Repeat for a few years and you have monitoring, of a sort: a pile of tripwires, each one commemorating a bad day.


What that loop can never produce is coverage of things that have not failed yet. The file server that has hummed along for four years has no checks, because it has never caused a postmortem. And it is also four years closer to the day it will. Meanwhile, the system that broke twice last quarter may have a dozen overlapping alerts, half of them noisy enough that people have quietly filtered them to a folder. Reactive monitoring overfits to the past. Strategy is the deliberate act of covering the future instead, and it starts with writing down what the monitoring is for before deciding what the monitoring is.


Meridian Parts ran exactly this kind of monitoring for six years. Every check on the system could be traced back to a ticket, and most of them to a particular bad week. The license server for the quoting software had never failed, so it had never earned a check. When it finally stopped the first alert was a sales rep standing in the server-room doorway. The incident review recorded the root cause as "not monitored" and the corrective action as "add monitoring". That was how every previous check had arrived as well. The admin wrote the goals page the following Monday. It took twenty minutes, which was shorter than the review meeting.


The three questions to answer in writing


All three questions can be answered without owning a single tool, and they should be. Answer them in writing, because writing forces choices. A hallway conversation produces nods, while a page produces commitments that survive staff turnover, budget season, and your own fading memory of why anything was set up.


Question 1: What does the business actually need to keep working?

Answer this one in business language, not device names. "People can take orders." "The warehouse can print shipping labels." "Staff can sign in and open their files." "Customers can reach us by phone." Not "the ERP server" — ERP is enterprise resource planning, the application the orders live in — not "switch stack 2," not a single IP address. The distinction matters because devices are means and services are ends. A server deserves attention exactly in proportion to the services standing on top of it. You cannot judge that proportion until the services have names. A list of five to ten such services is normal for a small or mid-size company. The test for each entry is simple: would a department head recognize it as a thing their team does? If a line item only makes sense to you, it is an ingredient, not a service — push it down a level and ask what it is for.


Question 2: How fast must we find out when each one breaks?

This is your detection goal: a promise about how quickly you will notice a failure. That is a different promise from how quickly you will fix it. Set one per service, and set it honestly. If you are a one- or two-person team, "we detect order-system failures within five minutes, around the clock" is a fantasy. The exception is when a machine is doing the detecting and something loud is doing the waking. And even then, five minutes at 3 AM only matters if someone will actually act at 3 AM. It is perfectly respectable for a small shop to promise five minutes during business hours and thirty overnight for the crown jewels. Next-business-day is respectable for things nobody uses after 6 PM. Detection goals connect directly to the uptime promises you can realistically make. The guide to setting internal uptime targets covers that arithmetic. The honest version of both documents is worth far more than the impressive version.


Question 3: Who needs to be told what?

Failures have audiences, and each audience needs different information. At minimum there are three. You (and any teammates) need full technical detail, immediately, on a channel you actually notice. Your boss needs the service-level story in plain words — what is down, who feels it, when the next update comes — and usually only for visible or lengthy outages. The affected users need to know that the problem is known, whether a workaround exists, and when to expect news. What they must not get is a copy of your technical alerts. Writing this down now, per service, prevents the two classic failures later. There is the boss who learns about an outage from a customer, and there is the all-staff email full of hexadecimal.


The one-page scope statement


Your three answers now condense into a single page called a monitoring scope statement. It contains the list of services you monitor, how fast you commit to detecting their failures, and who gets told. And — crucially — it states what you have decided not to monitor. Here is a skeleton you can copy into any text file:


MONITORING SCOPE STATEMENT — (company)      last reviewed: (date)

1. SERVICES WE MONITOR
   (service, in business words) — detect within (time),
   business hours / around the clock — on failure tell (who)
   ... one line per service ...

2. EXPLICITLY NOT MONITORED
   (system or area) — because (reason) — revisit (date/trigger)
   ... one line per exclusion ...

3. NOTIFICATION MAP
   Admin team: all alerts, immediately, (channel)
   Management: user-visible outages over (duration), plain language
   Users/departments: outages affecting them, with workaround + ETA

4. NEXT REVIEW: (date)        OWNER: (name)

One page is not a simplification you will outgrow — it is the point. A page can be re-read before every change, pinned to the wall, handed to a new boss, and actually kept current. Anything longer becomes a document about monitoring instead of a document that governs it. A long strategy document has one ambition in life, which is to be filed. Details like per-device thresholds, check frequencies, and tool settings belong a level down, in the requirements document and in the tool itself — not here.


Why the exclusions are the powerful part


Section 2 looks like an afterthought. It is the most valuable ink on the page. An explicit exclusions list does three jobs that nothing else in your monitoring setup can do.


It stops scope creep. Monitoring requests arrive constantly. For example, a department wants their pet dashboard watched. A vendor suggests an agent on everything. You feel vaguely guilty about the test lab. Without a written boundary, each request is a fresh negotiation and the default answer drifts toward yes, one check at a time, until the system is unmaintainable. With the list, the answer is: "that's on the exclusions list — here's why, and here's what it would take to move it up." The conversation shifts from guilt to trade-offs.


It prevents blame. When an unmonitored system eventually breaks — and it will — one thing separates a career-bruising meeting from a calm one. Was the non-coverage a decision or an accident? "The showroom signage player was on the exclusions list you approved in the spring review" is a sentence that ends discussions. "We never got around to it" starts them.


It forces honest conversations. Presenting the exclusions to management does something subtle: it makes non-coverage their decision too. Either they accept the exclusion, or they discover they care enough to fund the time and licensing to change it. Both outcomes are wins for you. An exclusion is never forever — each line carries a reason and a revisit trigger — but while it stands, it stands in writing.


Tip: write exclusion reasons a future stranger could evaluate, not just "low priority." Good reasons look like: "vendor-hosted, no visibility possible from our side," "scheduled for decommission," or "failure is annoying but nobody's work stops." A reason that no longer holds is exactly what the revisit date is for.


Stakeholder interviews for a small org


Here is the uncomfortable truth about Question 1: you cannot answer it from the server room. You know what the infrastructure looks like; you do not fully know what the business does with it. And every admin who thinks otherwise is running on assumptions that are at least two reorganizations old. The fix is cheap: fifteen minutes with each department head or team lead. You do not need a survey tool or a workshop — you need four questions and a notepad:


  • 1. Walk me through a normal day for your team. Not their systems — their work. You are listening for the actions: taking orders, printing labels, invoicing, answering calls.
  • 2. If something stopped working right now, what would actually halt that work? This surfaces the true dependencies, in their words.
  • 3. What breaks most often, and what do you do when it does? The workarounds people describe are a map of your existing unreliability — and of what they have stopped bothering to report.
  • 4. If something had to be down for a whole day, what could you live with? This is the priority question asked backward, and it gets far more truthful answers than "what's critical?", to which everyone answers "everything."

Then comes the translation step, because people describe pain, not services. Your notepad will say things like "the system is slow every morning" and "we can't work when the internet is down." Your job is to convert each feeling into a named service and, later, a check. (The CRM in the first row is the customer relationship management system, where sales keeps every contact and quote.)


What they sayWhat you write down
"The internet is always down"Cloud CRM reachable and responsive from the office
"Everything is slow first thing"Sign-in and file-open times during the 8–9 AM peak
"The warehouse thing crashes"Label printing at the packing stations
"Phones cut out sometimes"Call quality on the voice system, not just phone uptime

Expect two surprises; the interviews produce them every single time. The first is the critical system nobody mentioned to IT. That might be the label printer that gates every shipment, the spreadsheet macro the whole quoting process runs through, or the ancient license server for the design software. It has no redundancy and no checks, precisely because it never made it onto an official list. The second surprise is the mirror image: the beloved system nobody actually uses. The intranet portal everyone insisted was vital turns out to have been broken for nine days in the spring without a ticket. Both discoveries reshape your service list more than any technical audit would, and both are why the interviews are not optional.


From goals to a check philosophy


With goals on paper, something changes about how checks get chosen. The tool-driven approach — the default when there are no goals — is to enable whatever the tool can see. Discovery finds four hundred interfaces, so four hundred interfaces get monitored, because the checkbox was right there. The goal-driven approach flips the direction: a check exists because a goal needs it, and for no other reason.


The working test is traceability in both directions. Point at any check and you should be able to name the service goal it protects. Point at any goal and you should be able to list the checks that would actually catch its failures. A check with no goal is noise waiting to happen — retire it. A goal with no checks is an outage waiting to be discovered by a user — that gap is your next piece of work. You will not keep this perfectly true week to week. That is why a recurring coverage review exists to walk the two directions on a schedule. And when you sit down to build the first checks for each goal, you do not need to invent them from scratch. The checks in the ten essential checks map onto the goals almost every small business writes down.


Three anti-patterns to avoid


Monitoring what is easy instead of what matters

Ping answers cheaply, so everything gets pinged. The test VM (virtual machine) speaks every management protocol, so it wears forty sensors. The label printer that gates all shipping is awkward to query, so it gets nothing. Effort flows toward the cooperative devices instead of the important ones. The goals page is the antidote: importance was decided in Question 1, before anyone knew which devices would be easy. If the crown-jewel service depends on something hard to monitor, that difficulty is your problem to solve — not a reason to quietly watch something else instead.


Copying someone else's estate

Template libraries, a friend's exported config, the "best practice" checklist from a vendor forum — all tempting, all shaped by someone else's goals. Their compliance rules, their staffing, their tolerance for noise, and the one admin who understood every threshold, who did not come with the export. Imported wholesale, another company's monitoring gives you their blind spots plus a thousand alerts calibrated to a business you do not run. Borrow mechanics freely — threshold ideas, check techniques — but never borrow scope. Scope is the part that has to be yours.


Tool-first shopping

The most expensive anti-pattern: choosing the software before writing the goals. Without your own list, the requirements come from the glossiest feature matrix in the room. You end up comparing tools against each other instead of against your needs. That is the RFP trap, named for the request-for-proposal document, where the winner is whoever demos best rather than whatever fits. Buy nothing, trial nothing, sit through no demo until the goals page exists. It is one page. It takes twenty minutes and converts every vendor conversation from "look what it can do" to "here is what we need — show me." The full shopping discipline gets its own guide in the requirements document.


A worked example: 120 people, one page


Here is the whole process, start to finish, for a fictional but typical company: a 120-person building-products distributor. One office, one warehouse, a two-person IT team, most software still on premises, email in the cloud. The admin interviews six department heads over two days — sales, warehouse, finance, purchasing, customer service, and the general manager.


The interviews produce the two ritual surprises. The critical system nobody mentioned is the pair of thermal label printers in the warehouse. Every single outgoing pallet passes through them — no labels, no shipping. And finance adds that late shipments trigger contract penalties with the two biggest customers. The beloved ghost is the internal news portal, defended passionately by two managers. The web logs show it was visited eleven times in the last month, nine of them by its author. The service list gets rewritten accordingly, and the one-page statement comes out like this:


Service we monitorDetect withinCoverageOn failure, tell
Orders can be entered (ERP app + database)5 min7 AM–7 PM; 30 min overnightIT phone; GM if >30 min
Warehouse can print labels10 minShift hours, 6 AM–6 PMIT + warehouse supervisor
Email flows (cloud tenant reachable)15 minAround the clockIT; GM if >1 hour
Staff can sign in and open files10 minBusiness hoursIT; affected departments
Customers can call us (phone system)10 minBusiness hoursIT + customer service lead
Payroll system available4 hours30 min during the 3 run days/monthIT + finance

Then come the exclusions, each with a reason and a revisit trigger. The test lab is excluded (nothing production touches it; revisit if that changes). So are personal desk printers (failure annoys, never halts; the warehouse label printers are emphatically not in this line). The marketing analytics site is excluded (vendor-hosted, no visibility from our side; revisit at contract renewal). So is the boardroom AV kit (the projector and conferencing gear, checked manually before board days). The news portal is excluded too (pending a usage review with its two defenders). The notification map follows the pattern from Question 3. The review date is set six months out. And the admin's name goes in the owner box. Total length: one page. Total writing time: about twenty minutes, on top of two days of pleasant conversations that also bought the IT team more goodwill than a year of ticket-closing.


Where to go next


That page now governs everything downstream: which checks exist, which alerts wake someone, what the dashboards show, and what you can decline with a clear conscience. Twenty minutes of writing steers years of work, and I know of no better ratio anywhere in IT. Nothing built on it is archaeology, because every check will have been put there on purpose and the next admin can read why. The next steps in this pillar build directly on it. They include turning goals into a requirements document that makes tool decisions rational. They also include mapping each service to its dependencies so every line on your page becomes concrete checks. For the wider question of where monitoring boundaries belong, see Monitoring Scope: What to Watch (and Skip). For the rest of the strategy series, head back to Building a Monitoring Strategy.


Put this into practice with SysPeep Monitor

The goals page is deliberately tool-independent — but once it exists, you need software that can execute it. SysPeep Monitor discovers the devices behind your services automatically and watches them over SNMP, WMI and SSH. It lets you shape thresholds, dependencies and alert rules to match the detection goals you just wrote down. The free trial installs on any Windows machine in a few minutes.

Download Free TrialLearn About SysPeep