Acutis Network or machine, across 600 computers
The short answer: the manual method works. Compare two devices, split the path, ping the gateway then the internet — that settles one laptop in about a minute, and it is still the right thing to do.
It breaks at scale for a reason that has nothing to do with the technique: it only runs when someone complains, on one machine, after the fact. A fleet's real problem is not diagnosing the ticket in front of you. It is the tickets that never get filed, the ones you can't reproduce because the moment has passed, and the ones you can't tell apart because each was investigated alone.
The four places the manual method quietly fails
1. It's reactive, so you only see what gets reported
Most people do not file a ticket when the connection is bad for ten minutes. They reload, wait, blame the internet, and get on with their day. So the events you investigate are a biased sample: the ones bad enough or long enough to be worth a complaint. If a site has been degraded every afternoon for three weeks, the ticket queue may contain nothing about it at all.
2. The evidence is gone by the time you look
Someone reports that things were slow an hour ago. You run the tests now and everything is fine. There is no honest verdict to give — not because you lack skill, but because the fault is not happening while you are measuring. Intermittent problems are the majority of real network complaints and the manual method is structurally blind to them.
3. One machine can't tell you which side is at fault
This is the important one. On a single laptop, "compare against another device" is the whole trick — and at your desk you have a phone to compare against. Across a fleet you have the same question with no comparison available: a user in one office says the VPN is slow, and you have no idea whether the other forty machines at that site agree. The signal that decides network-versus-machine is the pattern across many machines at once, and that is exactly the thing one-at-a-time triage cannot see.
4. Nothing accumulates
Every investigation starts from nothing and ends when the ticket closes. Ask "is this getting worse?" or "did the change last month help?" or "which site is actually our problem?" and there is no record to answer from. Ninety separate one-off diagnoses do not add up to knowing anything about the fleet.
What it costs, concretely
| Situation | One machine | A fleet |
|---|---|---|
| Something is slow | Run two tests, compare, done | No idea if it's one user or a whole site |
| Problem is intermittent | Wait and retest | Evidence gone; ticket closed "could not reproduce" |
| User blames the network | Prove it in a minute | Tech dispatched to a machine fault |
| Network really is at fault | Call the ISP | Weeks of individual tickets before the pattern is spotted |
| "Is it getting worse?" | You remember | Unanswerable — nothing was recorded |
The expensive failure is the third row. A machine fault reported as a network fault sends someone to the wrong place, and a network fault dismissed as "that user's laptop again" sits unfixed while more tickets pile up behind it. Both are misattribution, and misattribution — not diagnosis time — is where the real cost sits.
What actually changes at scale
The fix is not a better test. The tests are fine. What changes is when and how many:
- Continuous instead of on-complaint
- The checks run all the time, so the intermittent afternoon problem is recorded while it is happening rather than reconstructed afterwards from memory.
- Many machines instead of one
- Ten machines at one site degrading together is a network fault, stated as a fact rather than a hunch. One machine degrading while its neighbours are fine is a machine fault. That comparison is free once every machine reports.
- History instead of a moment
- "This site has been bad every weekday afternoon for three weeks" is a sentence you can take to an ISP or a vendor. A single screenshot is not.
None of that requires touching the network. It requires the endpoints to answer the same question they always answered — is it the network or the machine? — continuously, and in one place.
That is what Acutis Go does
A free 7 MB observe-only agent on Windows, macOS and Linux runs the same checks every 60 seconds and reports a plain verdict — network or machine — for every machine, with the history behind it. It is observe-only by design: it changes nothing, opens no inbound port, and cannot push configuration, so the security review is short. Start with one machine free.
Get the agent