Building a One-Page Network Health Scorecard
Your monitoring system collects thousands of numbers a day. The people who decide your budget read exactly none of them. When the boss asks "how's the network?", the honest-but-useless answer is "fine, mostly." And so the network stays invisible until the day it breaks. That is the worst possible moment to become visible. There is a fix, and it is almost embarrassingly low-tech: a scorecard. One page. Once a month. Six to ten lines a non-engineer can read in two minutes.
This guide shows you how to build one from data you already have. It covers how to choose the lines and define each measurement precisely enough that next month's number is genuinely comparable to this month's. It also covers how to set simple traffic-light thresholds and present the whole thing honestly. You get a worked example filled in and a blank template to copy. And — the quiet payoff — you get a tool that changes how budget conversations go.
In this guide
- What a scorecard is (and is not)
- Choosing your six to ten lines
- Defining each line precisely
- Traffic-light thresholds that mean something
- A worked example: one month, one page
- Where each number comes from
- Presenting honestly
- The budget conversation
- Your blank template
What a scorecard is (and is not)
A network health scorecard answers three questions for a reader who will never open your monitoring console: Is the network doing its job? Is it getting better or worse? Is anything about to bite us? That is the entire mission. It is not a dashboard — dashboards are live instruments for you, the operator. It is not a report dump — forty pages of graphs is a doorstop, not communication. It is one page, in the same format every month, written in units a manager already understands: percentages, minutes, and colors.
The design principle behind everything in this guide is comparability beats sophistication. Think of a bathroom scale: a cheap one used every morning, same time, same place, tells you more than a laboratory instrument used randomly. A scorecard is a cheap scale for your network. The individual numbers matter less than the fact that they are produced the same way, month after month. That makes the trend real. Every decision below — which lines, which definitions, which thresholds — serves that principle.
Choosing your six to ten lines
A good scorecard line passes three tests. You can measure it from data you already collect. It maps to something users actually feel. And its definition can stay stable for years. Here is the candidate menu — the first seven make a strong starter set for almost any small or mid-size network:
- Internet availability %. The share of the month your internet link actually worked. The single most-felt number in the building; if you have two links, give each its own line.
- Latency to the outside world: mean and 95th percentile. Latency is round-trip delay. The mean is the typical experience. The 95th percentile is the value 95% of samples stayed at or below. In plain words, that is "how bad a normal bad moment gets." Report both: the pair exposes a network that is usually fine but regularly awful, which an average alone hides.
- Packet loss %. The share of test packets that never arrived. Small numbers matter here: even one percent ruins voice and video calls.
- Busiest-link peak utilization. How full your most-loaded uplink runs at peak, measured as a 95th percentile rather than a one-off spike. This is your capacity early-warning line — the one that eventually justifies upgrades.
- Wi-Fi health. Simplest robust version: total minutes any access point was down this month. Users experience a dead AP as "the network is broken," so the scorecard should too.
- User-facing incidents. A count of network-caused problems that affected at least one real user.
- User-facing downtime minutes. The sum of how long those incidents actually stopped people working. Count and minutes together tell the story: five two-minute blips is a different disease than one two-hour outage.
Sensible extras once the core is running: VPN availability if remote work matters, DNS response health, or availability of one business-critical internal service. And two things to leave off: vanity numbers (total gigabytes moved is interesting trivia, not health), and anything you cannot reproduce next month from recorded data. If a line's number would come from memory or vibes, it does not belong on the page.
Why the six-to-ten band? Below six, the page usually means you have merged things the reader needs to see separately. Take incidents and downtime, say, which move independently and mean different things. Above ten, the page stops being readable in two minutes. It starts becoming the forty-page report nobody opens, with lines filled from habit rather than attention. The one-page constraint is not a limitation to work around. It is the feature. It forces on you exactly the prioritization that makes the result worth reading.
Defining each line precisely
Here is where most scorecards quietly die. "Availability" sounds self-explanatory until you ask: measured from where? To what? How often? Do maintenance windows count? If the answers drift from month to month, your trend line is fiction. So every line gets a written definition with five parts. They are the source (which system measures), the target, the frequency, the formula, and the inclusion rules (what counts, what gets footnoted). Like this:
| Line | Example precise definition |
|---|---|
| Internet availability | % of one-minute ping checks from the monitoring server to the ISP gateway that received a reply. All hours count; maintenance is footnoted, never removed. |
| Latency, mean / 95th | Mean and 95th percentile of round-trip time from those same checks, in ms, across the whole month. |
| Packet loss | Unanswered pings as % of pings sent, excluding minutes already counted as outage (so one problem is not billed twice). |
| Busiest-link utilization | 95th percentile of five-minute utilization samples on whichever monitored uplink scored highest this month; the link is named in Notes. |
| Wi-Fi health | Minutes any access point was down, summed across all APs, from monitoring up/down history. |
| User-facing incidents | Count of network-caused issues affecting at least one user, from tickets plus the outage log. |
| User-facing downtime | Minutes users could not work due to network causes, summed per incident from the outage log. |
Availability deserves one extra piece of arithmetic, because percentages near 100 hide their meaning. Translate them into minutes for a 30-day month:
30-day month = 43,200 minutes 99.9% available = about 43 minutes down 99.5% available = about 3.6 hours down 99.99% available = about 4 minutes down
Put the minutes in parentheses right on the scorecard — "99.88% (51 min out)" — and the line explains itself. Nobody has intuition for the gap between 99.9 and 99.5. Everyone has intuition for 43 minutes versus three and a half hours.
Traffic-light thresholds that mean something
Raw numbers still force the reader to ask "is that good?" — so answer it for them with three colors per line. Green means the network did its job. Amber means nothing is broken, but watch this or plan spending. Red means users were hurt and action is happening. The colors carry the judgment so the reader does not need to know whether 0.4% loss is bad. (You know; that is the service you are providing.)
Set thresholds from your own baseline — your first two or three months of real data — not from wishes or vendor brochures. If your internet has delivered 99.95% for three months, green at 99.9% is honest. In that case, green at 99.999% is theater. Here are reasonable starting points to tune against your baseline. Set availability green at or above 99.9%, red below 99.5%. Set loss green at or below 0.1%, red above 0.5%. Set busiest-link 95th percentile green below 70%, amber to 85%, red above. Two rules keep the colors trustworthy. Thresholds are part of the definition, so changing one gets a footnote. And amber should be slightly too easy to hit. A scorecard that ambers a month before trouble is doing its job. One that is green until the outage is decoration.
One more habit: audit the colors yearly. A line that has stayed green for a full year is either a quiet triumph or a wasted row. If nothing that line could plausibly report would change anyone's behavior, swap it for something you are genuinely uncertain about. The exception is availability, which keeps its permanent seat as the heartbeat of the page even when it is boringly green. Especially when it is boringly green; a long run of green availability is the credibility you will spend later.
A worked example: one month, one page
Here is a filled-in scorecard for a fictional but very recognizable small company — one office, two hundred people, one ISP link, a dozen access points:
| Scorecard line | This month | Last month | Status | Notes |
|---|---|---|---|---|
| Internet availability | 99.88% (51 min out) | 99.99% (4 min) | Amber | [1] [2] |
| Latency, mean / 95th | 12 / 19 ms | 12 / 18 ms | Green | |
| Packet loss | 0.1% | 0.1% | Green | |
| Busiest-link utilization (95th) | 78% (WAN out) | 71% (WAN out) | Amber | third rise in a row |
| Wi-Fi: AP-down minutes | 0 | 240 | Green | failed AP replaced |
| User-facing incidents | 2 | 1 | Green | [1] [2] |
| User-facing downtime | 51 min | 9 min | Amber | [1] |
Footnotes: [1] ISP outage on the 14th, 38 minutes, provider ticket reference kept on file. [2] Core switch reboot during the announced maintenance window on the 21st, 13 minutes — included in the numbers and footnoted, not silently removed.
Now read it the way your manager will. Sixty seconds tells them: the month's downtime traces almost entirely to one ISP incident, documented, with a ticket to wave at the provider. Wi-Fi got fixed and stayed fixed. And there is one quiet, important story. The busiest link has ambered three months running — 64%, 71%, 78%. That is not an emergency but is unmistakably a trajectory. That single row is doing more budget preparation than any speech you could give, and you did not have to say a word.
Where each number comes from
Everything on the example page falls out of ordinary monitoring data plus one text file:
- Availability, latency, loss: from the history of a scheduled ping check your monitoring system already runs against the ISP gateway. Most tools produce an uptime percentage and response-time statistics per sensor per month. If yours does not compute a 95th percentile, export the samples to a spreadsheet and use its percentile function.
- Busiest-link utilization: from the interface traffic your monitoring collects over SNMP. Check each uplink's monthly 95th percentile and report the highest. One classic trap: make sure the tool polls 64-bit counters on fast links. Otherwise, a counter wrap can hand your scorecard an impossible number. The SNMP Monitoring guides cover this from scratch.
- Wi-Fi health: from AP up/down alert history, summed — or your controller's uptime report if you have one.
- Incidents and downtime: from tickets tagged as network-caused, cross-checked against your outage log.
Tip: start the outage log today. Use a plain text file with one line per incident. Include the date, start time, end time, what broke, what fixed it, roughly how many users were affected. It costs two minutes per incident and it is the raw material for your two most persuasive scorecard lines. Reconstructing the same facts from memory at month-end is somewhere between painful and fictional.
Budget the first assembly honestly: expect an hour, mostly spent finding where your tool hides its monthly statistics. By the third month it is a twenty-minute routine. Put a recurring appointment on the first business day of the month. A scorecard that ships erratically loses its authority fast.
Presenting honestly
A scorecard's entire value is that people trust it, and trust has simple mechanics. Footnote, never delete. Maintenance windows and one-off disasters stay in the numbers, marked and explained. The reader learns the month as it happened, and learns that your numbers do not flinch. Never restate a published month. If you find a measurement error, correct it visibly with a note, the way accountants do. Mark definition changes loudly. If a line's recipe changes, say so on the page and treat the trend as broken at that point. Comparing new-recipe numbers to old-recipe numbers is fiction with extra steps.
And publish the bad months with the same fonts as the good ones. A scorecard that is always green reads as propaganda. A reader who suspects propaganda discounts everything on the page — including the greens you earned. Notes should explain, not excuse. "38 minutes, ISP fault, ticket filed and credit requested" is information. A paragraph about how it was not your fault is noise. The tone to aim for is a lab notebook, not a press release.
Finally, keep the raw data retrievable. The scorecard is a summary of measurements, not a substitute for them. So file the monthly exports and graphs behind each page where you can actually find them. Eventually someone will ask, "where does this 78% come from?" Producing the underlying interface graph in one minute is what separates an instrument from an opinion.
Warning: the first time someone catches a cherry-picked number, every future scorecard you publish will be read as marketing. Cherry-picking might mean a bad week quietly excluded, or a threshold nudged after the fact to keep a line green. There is no recovering the discount rate. Publish the ugly month; the ugly month is usually your best budget evidence anyway.
The budget conversation
Here is what the scorecard actually changes. Without one, capacity requests sound like adjectives: "the internet feels slow, we should upgrade." Adjectives lose to other departments' spreadsheets every time. With a scorecard, the same request sounds like arithmetic: "the busiest link's 95th percentile has gone 64, 71, 78 over three months. On this trajectory we hit saturation. And here is what saturation did to the branch office last time." Trends get funded; feelings get deferred.
Downtime minutes convert into a currency managers already trade in. Fifty-one minutes of outage affecting forty people is thirty-four lost user-hours. You provide the minutes and the multiplication, and let the reader attach their own money to it. That is more convincing than any figure you could assert. The availability line, kept honestly for a year, also becomes leverage of a different kind. It is evidence when your ISP's delivery drifts below what the contract — the SLA, or service level agreement — promises. It is also the paper trail behind credit requests or a switch of provider.
The scorecard also protects you in both directions. When a warning you raised was declined, the amber months are timestamped. That is not for blame, but so the next conversation starts from shared facts. And when the data says the fashionable upgrade is not needed, the scorecard says that too. Saying it costs you nothing but buys enormous credibility for the next time you ask. An instrument that sometimes argues against its owner is an instrument people believe.
Your blank template
Copy this table, swap in your own lines, and staple your written definitions underneath it. The definitions table above is the other half of the template. Keep the Status column to the three words Green, Amber, Red. The words survive black-and-white printing and colorblind readers alike:
| Scorecard line | This month | Last month | Status | Notes |
|---|---|---|---|---|
| Internet availability | ||||
| Latency, mean / 95th | ||||
| Packet loss | ||||
| Busiest-link utilization (95th) | ||||
| Wi-Fi: AP-down minutes | ||||
| User-facing incidents | ||||
| User-facing downtime (min) | ||||
| (your line) |
The last-month column is the smallest trend you can print. If the page has room, a third number column makes slow drifts visible that two adjacent months hide. That column could show three months ago, or the same month last year on networks with seasonal rhythms.
And the adoption sequence, start to steady-state:
- 1. Pick your lines — the starter seven above, adjusted to your network.
- 2. Write the precise definition for each: source, target, frequency, formula, inclusion rules.
- 3. Pull two or three months of history from your monitoring to establish the baseline.
- 4. Set green/amber/red from that baseline, slightly strict on amber.
- 5. Fill in the first page, footnotes and all, and publish it — imperfect but on time.
- 6. Repeat on the same day every month; review thresholds once a year, with a footnote when they move.
Where to go next
A one-page scorecard is the rare project that is both easy and career-changing. It turns your monitoring data into a story non-engineers can follow. It catches slow-building problems while they are still cheap. And it replaces "trust me" with a trend line. The quality of the page rests on the quality of the measurements underneath it. That is what the rest of The Core Network Metrics teaches. It covers what each metric means, how to measure it with free tools, and how to avoid the classic misreadings. The wider topics library covers the monitoring craft around it.
Put this into practice with SysPeep Monitor
Nearly every number on the scorecard falls out of monitoring history — if the history exists. SysPeep Monitor collects availability, response times, and interface utilization over SNMP and WMI around the clock. Its reports pull the monthly numbers from a single web console. So assembling the page takes minutes instead of an archaeology session. The free trial installs on any Windows machine.
Download Free TrialLearn About SysPeep