The Blind-Spot Audit: A Self-Assessment With Scoring Sheet
Reading about blind spots is one thing. Knowing which ones your network has, ranked, scored, and written down, is another. It is the difference between a vague feeling that "we should have better visibility" and a plan you can start on Monday. This guide turns the whole subject into an afternoon's exercise. It provides a worksheet of concrete questions, a simple score, and a method for converting that score into your next three fixes.
The questions are deliberately answerable. None of them needs a consultant, a budget, or special tooling. Each can be settled by a junior admin looking at a screen, a closet, or their own honest memory. That is the point. A blind-spot audit is not a test of your equipment. It is a test of what you can currently see, and you find that out by trying to look.
A low score is the normal starting point. Almost everyone who runs this audit for the first time scores low, because networks grow organically and visibility never keeps up on its own. The first time I scored a network of my own I got fourteen, and I had built most of its monitoring. The score's job is not to grade you; it is to hand you a priority list. Treat it that way and the whole exercise stays useful instead of demoralizing.
In this guide
- How the audit works
- The worksheet: eight classes, 24 questions
- Scoring your audit and reading the bands
- The results summary sheet
- From score to plan: the first cheap move per class
- Cadence: repeat, and track the score
How the audit works
The audit walks the eight classes of blind spot surveyed in The Eight Places Your Monitoring Can't See — from the traffic that never crosses your firewall to the physical closets nobody visits. Each class gets three questions, 24 in all. Score every question 0, 1, or 2, where a 2 is an answer you would bet lunch on:
- 0 — no idea. You could not answer this today, and you are not sure how you would find out.
- 1 — partial. You could piece an answer together with some digging, or the answer is "sort of, for some of the network."
- 2 — confident. You can answer now, from data you already collect.
Three ground rules keep the result honest. First, answer by looking, not from memory. If a question asks whether you could name your top talkers, actually try to name them. The gap between "I think we can" and "we just did" is precisely what the audit measures. Second, score what exists today, not the project that is planned or the feature the vendor promised. Third, nobody is grading you. Do the audit alone or with your team, but keep it blameless — the enemy is the dark, not whoever inherited it.
Time required: an honest afternoon, blocked out in the calendar, because ticket queues are excellent at preventing audits. The run of show:
- Step 1: Print the tables (or copy them into a document) and block out three uninterrupted hours.
- Step 2: Work classes 1 through 7 from your desk with the relevant consoles open.
- Step 3: Do class 8 last — it involves standing up and visiting closets.
- Step 4: Fill in the summary sheet, pick your two lowest classes, and put the re-audit date on the calendar before the momentum fades.
The worksheet: eight classes, 24 questions
Before you start, open the screens you will need: the switch admin pages, the firewall console, the DHCP server, the VPN gateway, and your monitoring dashboard. Nearly every question below is answered by looking at one of those five. Circle one score per row, and jot a one-line note wherever you score 0 or 1 ("no flow data anywhere," "management VLAN has no backup path"). Those notes become your plan later. If you can, audit in pairs: one person drives the consoles while the other scores. It goes faster, and two people arguing over whether something deserves a 1 or a 2 is exactly the honesty mechanism this exercise needs. And do class 8 even though it involves standing up; that is exactly why it is the class most often skipped.
Class 1: East-west traffic
East-west traffic moves sideways between your internal systems — workstation to file server, server to database — and never crosses the firewall where most visibility lives. A "top talker" is simply the device moving the most data on a link.
| # | Question | Score |
|---|---|---|
| 1.1 | Could you name the top three talkers between your two busiest VLANs right now? | 0 / 1 / 2 |
| 1.2 | If a workstation started copying the entire file server tonight, would anything record it? | 0 / 1 / 2 |
| 1.3 | Do you keep traffic history for your switch-to-switch uplinks, or only for the internet link? | 0 / 1 / 2 |
Class 2: Encrypted traffic
You cannot read encrypted content — and mostly should not try — but volume, timing, and destinations remain visible. And they carry real diagnostic and security signal.
| # | Question | Score |
|---|---|---|
| 2.1 | For your busiest link, can you say who your servers talk to (destinations), even without content? | 0 / 1 / 2 |
| 2.2 | If a server began uploading unusually large volumes to an unfamiliar destination, would anything alert? | 0 / 1 / 2 |
| 2.3 | Do you record connection metadata (source, destination, volume, time) anywhere, for any segment? | 0 / 1 / 2 |
Class 3: Unmanaged and undocumented devices
Every port on a managed switch reports which hardware addresses (MACs) it sees. Two MACs on a port documented for one device means something undocumented is plugged in — often a cheap unmanaged switch under a desk.
| # | Question | Score |
|---|---|---|
| 3.1 | Does every access port's MAC count match expectations — one device, or a documented reason why more? | 0 / 1 / 2 |
| 3.2 | If you swept a subnet for live addresses right now, could you name every responder? | 0 / 1 / 2 |
| 3.3 | Do you know how many unmanaged ("dumb") switches exist on your network, and where they are? | 0 / 1 / 2 |
Class 4: Guest and IoT segments
The segments you deliberately walled off — guest Wi-Fi, cameras, sensors — still share your hardware and your internet connection, and their failures still generate tickets.
| # | Question | Score |
|---|---|---|
| 4.1 | If the guest Wi-Fi broke right now, would you learn it from a dashboard or from a visitor? | 0 / 1 / 2 |
| 4.2 | Do you watch the guest and IoT address pools (DHCP scopes) for exhaustion? | 0 / 1 / 2 |
| 4.3 | Could you spot a camera or sensor that tripled its normal traffic this week? | 0 / 1 / 2 |
Class 5: Remote sites and home offices
Branch sites and home workers live on networks you barely run or do not run at all. The question is not whether you can see their homes (you must not). It is whether you can see the work path, from the laptop to the concentrator. That is the box at your edge where every VPN tunnel lands.
| # | Question | Score |
|---|---|---|
| 5.1 | Can you see per-user VPN tunnel state and reconnect history, or only "the concentrator is up"? | 0 / 1 / 2 |
| 5.2 | When a remote worker says "everything is slow," can you separate their Wi-Fi, their ISP, and your VPN edge with evidence? | 0 / 1 / 2 |
| 5.3 | Does anything test your key apps from outside the office, the way remote users reach them? | 0 / 1 / 2 |
Class 6: Cloud egress and SaaS traffic
Much of the working day now happens in services you do not host. These include mail, files, and meetings, delivered as SaaS (software as a service, applications that run in a provider's cloud). From inside, all of it looks like one busy pipe to "the cloud," and every SaaS slowdown looks identical to a network problem until someone can prove otherwise.
| # | Question | Score |
|---|---|---|
| 6.1 | Could you name which cloud services consume the most of your internet link right now? | 0 / 1 / 2 |
| 6.2 | When a cloud app is slow, can you show with evidence whether the cause is your link, your DNS, or the provider? | 0 / 1 / 2 |
| 6.3 | Do scheduled checks watch reachability and response time of your critical cloud services from your own network? | 0 / 1 / 2 |
Class 7: The management plane
The management plane is the network you use to control the network. It includes switch admin interfaces, the management VLAN, and the credentials and paths that let you drive. Ideally it includes an out-of-band path, a way in that does not depend on the network it manages. Lose it and you are locked out of your own infrastructure while it burns.
| # | Question | Score |
|---|---|---|
| 7.1 | If the primary switch-management VLAN broke, do you have another way in — console cable, out-of-band path — that you have actually tested? | 0 / 1 / 2 |
| 7.2 | Would you be alerted if a device's management interface stopped answering, even while its traffic still flowed? | 0 / 1 / 2 |
| 7.3 | Would you notice an admin login to a switch or firewall that was not you or a colleague? | 0 / 1 / 2 |
Class 8: Layer 1 — the physical world
This layer includes cables, closets, power (including the UPS, the battery unit that carries the gear through a mains cut), and temperature. Here "the network" is actual objects that corrode, overheat, and get unplugged to make room for a vacuum cleaner. No dashboard sees a closet nobody visits.
| # | Question | Score |
|---|---|---|
| 8.1 | When did someone last physically inspect every network closet — and would they have spotted a new device or cable? | 0 / 1 / 2 |
| 8.2 | Do you monitor error counters on important links (the classic signature of a failing cable or transceiver)? | 0 / 1 / 2 |
| 8.3 | Do you monitor the environment where gear lives — temperature, UPS and battery state? | 0 / 1 / 2 |
Bonus question (not scored): who watches the watcher? If your monitoring system died tonight, how long until a human noticed? A daily "I am alive" signal whose absence someone would spot — a dead-man's switch — is the classic cheap fix. A quarterly alert-path fire drill (cause a deliberate failure, confirm the right person is told) keeps the whole apparatus honest.
That is the whole worksheet. If some questions annoyed you because the answer was so plainly "no idea" — good. Annoyance is the audit working. Write the 0 down and keep moving.
Scoring your audit
Add each class's three questions for a subtotal out of 6, then add the subtotals for a total out of 48. The bands:
| Total | Band | What it honestly means |
|---|---|---|
| 0–16 | Mostly blind | The normal starting point — most networks live here until someone runs this exercise |
| 17–33 | Partial visibility | The basics are watched; the sideways, remote, and physical corners are still dark |
| 34–48 | Good visibility | Rare and earned; the work shifts from building visibility to defending it |
If you scored 0–16: mostly blind
First, exhale. This band is where almost everyone starts, because visibility never grows on its own. Networks accumulate segments, closets, and remote users far faster than anyone instruments them. Nothing about this score says you are bad at your job; it says nobody ever handed you a map. You now have one, which puts you ahead of every admin who never looked. Your move is aggressive triage: take your two lowest-scoring classes, apply the first cheap move from the plan table below, and re-score in three months. Early points are cheap points — at this band, one afternoon of work can move the total by five.
Stonebridge Insurance scored eleven on its first audit, and most of the eleven came from the internet link, which was beautifully graphed. Question 3.2 took the longest, because a sweep of the main office subnet returned a hundred and sixty responders against an inventory of seventy-four. The admin running it spent the rest of the afternoon finding out what the other eighty-six were. They turned out to be, in order of quantity, desk phones nobody had listed, three vendor boxes, a shelf of lab machines, and a games console. The second audit, three months later, scored nineteen, with the inventory at a hundred and fifty-eight and two responders still unexplained. Those two were written on the sheet as "dark", which, the admin pointed out, was the first time anything on that network had been.
If you scored 17–33: partial visibility
The common middle. Typically this band means the north-south story is solid — internet link, firewall, key servers — while east-west traffic, remote users, and the physical layer stay dark. The cost shows up in incident length: problems inside a watched zone get diagnosed in minutes. Problems that cross a dark zone take hours. And you can usually name the last incident that proved it. Spend your effort where subtotals are lowest rather than polishing the classes you already score 5s and 6s in. A 1-to-4 improvement in a dark class shortens more outages than a 5-to-6 in a bright one.
If you scored 34–48: good visibility
Genuinely rare, and worth saying so out loud to whoever funds your tooling. Your risk now is decay: every new VLAN, vendor box, and remote hire is born unmonitored, so visibility rots quietly unless discovery, onboarding habits, and re-audits keep pace. Guard the score the way you built it. Let the unscored bonus question keep you humble, because an alert-path fire drill that fails is most informative exactly when everything else looks green.
Tip: resist the urge to round up. A 1 that should be a 0 does not make the network any more visible — it just hides the hole from the one person positioned to fix it. The audit only pays if the dark spots end up marked "dark."
The results summary sheet
Copy this, fill in your subtotals, and mark priority by the simple rule: lowest subtotal first, break ties by whichever class caused your most recent painful incident.
| Class | Max | Your score | Priority |
|---|---|---|---|
| 1. East-west traffic | 6 | ||
| 2. Encrypted traffic | 6 | ||
| 3. Unmanaged devices | 6 | ||
| 4. Guest / IoT segments | 6 | ||
| 5. Remote sites / home offices | 6 | ||
| 6. Cloud egress / SaaS | 6 | ||
| 7. Management plane | 6 | ||
| 8. Physical layer (L1) | 6 | ||
| Total | 48 |
From score to plan: the first cheap move
A score without a next step is just a sad number. The table gives each class's first cheap move, something achievable in a day or less with the gear you already own. It also gives what "done enough" looks like at small-organization scale. Done enough is a real standard: past it, further investment in that class usually returns less than a first move in a darker one.
| Low-scoring class | First cheap move | Done enough (small-org scale) |
|---|---|---|
| East-west traffic | Graph every switch uplink's traffic counters; see the east-west guide | History on all uplinks, plus top-talker sampling on the busiest internal path |
| Encrypted traffic | Start keeping the boundary metadata you already generate (firewall logs); see the encrypted-traffic guide | Volume-and-destination baselines for servers, with an alert on big deviations |
| Unmanaged devices | One afternoon sweep: MAC tables and live addresses reconciled against your docs; see the unmanaged-devices guide | Every port's MAC count explained; the sweep repeated on a schedule |
| Guest / IoT segments | Watch each segment's uplink counters and DHCP scope from the managed side; see the guest & IoT guide | A per-segment watchlist at the infrastructure layer, plus one guest-side probe |
| Remote / home offices | Turn on per-user tunnel visibility at your VPN edge; see the home-office guide | Symptom-to-layer diagnosis with evidence, disclosed to users |
| Cloud egress / SaaS | Add reachability and response-time checks for each critical cloud service, run from your own network | One "is it us or them?" graph per critical service |
| Management plane | Document and physically test one out-of-band way into each critical device | Management interfaces monitored; a second way in you have actually used |
| Physical layer (L1) | Walk every closet with a camera and a checklist; enable error-counter alerts | Error alerts on key links; temperature and UPS watched; closet walk on the calendar |
Notice that none of these first moves involve buying anything. Blind spots persist not because the fixes are expensive but because they are invisible — which is exactly the problem the audit just solved.
And resist fixing all eight classes at once. Two. Not three, however cheap the third looks on a Friday. Two at a time is the sustainable pace for a small team: enough to move the score visibly, few enough that each fix gets finished rather than merely started. Everything else stays on the sheet, safely written down. A documented blind spot you have chosen to accept for now is a decision; an undocumented one is a surprise on layaway.
Cadence: repeat, and track the score
Run the audit twice a year, and additionally after anything that reshapes the network. That could be a new site, an office move, a VLAN redesign, a VPN replacement, or a merger's worth of inherited equipment. Big changes mint new blind spots faster than anything else, because every project plan ends at "it works" and almost none end at "and we can see it."
Keep the filled-in sheets and date them, because the running total becomes something rare in IT. It is a metric that only moves up when you pay attention, and that visibly rewards unglamorous work. Most infrastructure effort is invisible when it succeeds — nobody celebrates the outage that did not happen. A score that climbs from 14 to 22 to 31 across three audits is that work made visible, to you, your team, and whoever approves your time. It also makes an honest status report: "we are at 31 of 48, and this quarter we are buying points in the management plane" is a sentence a non-technical manager can both understand and fund.
Where to go next
Visibility compounds. Every blind spot you close does not just prevent its own class of surprise. It makes every future incident shorter, because diagnosis is elimination. Each zone you can see is a suspect you can clear in seconds instead of hours. The audit gives you the order to close them in. The rest of the Network Blind Spots pillar gives you the how, one dark corner at a time. Run the sheet, pick your two lowest classes, and go turn on some lights. Fourteen is not a bad number to start from. It is only the first one.
Put this into practice with SysPeep Monitor
A surprising number of these questions get easier the day basic monitoring is in place. SysPeep Monitor auto-discovers the devices and interfaces on your network. That by itself moves several unmanaged-device answers toward a 2. SysPeep Monitor then watches health, port status, and traffic counters over SNMP and WMI from one web console, with alerts when anything drifts. The free trial is a quick way to buy your score a few points this month.
Download Free TrialLearn About SysPeep