The characteristic small business network problem is not an outage. It is the thing that happens every Tuesday afternoon, that nobody can reproduce, and that the internet provider says is not their fault.
Monitoring is how that becomes a fault somebody can fix.
What to watch
Internet quality, continuously
Not just whether it is up. Latency and packet loss, recorded over time.
This is the single most valuable thing to monitor, because it produces evidence. A graph showing five per cent packet loss every weekday between two and four is something a provider will act on. "It sometimes goes funny in the afternoons" is not.
Run it from inside your network, continuously, to a reliable external target.
Whether key devices are reachable
The firewall, the switches, the access points, any server, the printer everybody uses, the till system.
Simple availability checking tells you within a minute that something has stopped, and more importantly tells you what has not — which narrows a problem immediately.
Switch port errors
Overlooked and diagnostically valuable. Switches count corrupted and discarded frames per port.
A port with a rising error count almost always has a physical problem: a damaged cable, a poor termination, a duplex mismatch, or a failing device. These are exactly the faults that produce intermittent symptoms, and the error counter finds them before anyone notices.
A monthly look at error counters across your switches will find problems nobody had reported.
Wireless health
How many devices are connected to each access point, and whether any are struggling. An access point with forty clients while its neighbour has three tells you something about placement — see office wi-fi.
Also worth watching: repeated authentication failures, which can indicate a device with stale credentials or something less benign.
Bandwidth use, by direction
Enough to answer "is the connection saturated?" and "by what?". A backup running at the wrong time of day is a common finding, and moving it fixes the daily slowdown nobody could explain.
Telling a real problem from a blip
Networks have transient events constantly. Monitoring that alerts on every one is monitoring nobody reads.
Require duration. Three consecutive failed checks rather than one.
Look at trends, not moments. A single latency spike is nothing. Latency that has been climbing for three weeks is a problem approaching.
Correlate. If everything reports a problem at once, look at the common element — usually the internet connection or the power. If one device reports a problem, look at that device.
Establish what normal is. Monitoring is most useful after a few weeks of baseline. Without it you cannot tell whether today's figure is unusual.
The value of monitoring is not knowing something is broken now. It is being able to say what changed, and when, three weeks after it started.
What to do with the information
With a provider. Evidence changes the conversation entirely. Timestamps, loss percentages and a pattern get a fault accepted where a description does not.
With capacity. If bandwidth is regularly saturated at particular times, you can decide whether to upgrade or to move whatever is consuming it.
With ageing equipment. Rising error rates on a switch, or an access point that reboots weekly, are arguments for replacement supported by evidence rather than by a hunch.
With planning. Growth in device counts and bandwidth tells you when the current arrangement will run out, before it does.
Keeping it proportionate
A small business does not need an enterprise monitoring platform. What it needs is:
- Continuous internet quality measurement
- Availability checking for a dozen or so key devices
- Access to switch statistics when investigating
- Alerts to a person who will act, not to a mailbox
- Enough history to see a trend — three months is usually plenty
That is achievable modestly, and it is normally included in a managed support arrangement. The wider principles are in monitoring and alerting basics.
The mistake to avoid
Installing monitoring, admiring the dashboard for a fortnight, and never looking again.
Monitoring is useful when something is reviewed regularly — a short weekly summary read by somebody, and alerts that reach a person out of hours. Without that, it is a system quietly recording evidence nobody consults.
Our maintenance team monitors business networks across Suffolk and Norfolk, with alerts to named people and a monthly summary in plain English. Start a conversation.
