Where the device count is small, the traffic is one kind, and nobody will ever need to diagnose it from somewhere else.

Good cases are genuinely common: a printer, a desktop and a scanner sharing one outlet; a machine and its HMI inside a control panel; a few devices in a remote building fed by one uplink. In all of them the switch is a port multiplier and nothing more.

The qualifying conditions are worth stating plainly. Everything on it belongs to the same network segment, so there is nothing to separate. Nobody needs statistics from it. And there is exactly one path out, so a loop cannot form through it.

Where any of those stop being true - a second uplink, a camera that should be on its own VLAN, a fault that has to be found without a site visit - the unmanaged switch is the wrong device and usually the thing that turns a small problem into a long one.

Visibility, segmentation, protection and control - four separate things, and the first is the one that hurts.

Visibility: no port statistics, no error counters, no traffic rates, no MAC address table you can read, no logs. A port with a failing cable shows as 'slow', with nothing to confirm it.

Segmentation: no VLANs. Everything connected shares one broadcast domain, so cameras, controls and business traffic cannot be separated on shared cabling.

Protection: no spanning tree, so a loop is not blocked; no storm control, so a broadcast storm is passed on; no port security, so any device plugged in is on the network.

Control: no ability to disable a port, set a speed, prioritise traffic, or mirror a port for analysis.

None of these matter on three devices behind a desk. All of them matter the moment the switch is part of a system somebody is responsible for.

It cannot create one by itself, but it will happily pass one on - and unlike a managed switch it will not stop it.

A loop needs two paths between the same points. The usual cause is human: a patch lead plugged into two wall outlets that both lead back to the same cabinet, or a second uplink added to a small switch that already has one.

A managed switch runs spanning tree, notices the redundant path and blocks it before any harm is done. An unmanaged switch has no such mechanism. Broadcast frames circulate, multiply and saturate every link in the segment within seconds.

What makes it particularly unpleasant is that the loop can be created at the desk while the consequence appears across the whole network, and there is nothing in the unmanaged switch to indicate where it is.

Some low-cost switches now include a simple loop detection that disables a port. It is worth looking for if unmanaged switches are going to be used in a shared cabling system at all.

Yes, and PoE is one of the few features that appears on unmanaged units without complicating them - although the absence of per-port control does have consequences.

Unmanaged PoE switches supply power automatically to any device that negotiates for it, up to the port and total budget. For a couple of cameras or an access point in a remote position, that is exactly what is wanted and needs no setup.

What is missing is control. You cannot check how much power a port is drawing, cannot set a priority so that the important device keeps power when the budget is exhausted, and - most usefully - cannot power-cycle a port remotely to reboot a frozen camera.

That last point converts a five-minute remote action into a site visit, and on remote or rooftop devices it is often the single argument that justifies a managed switch.

Check the total budget rather than the per-port figure, since unmanaged units are frequently specified light.

Effectively none as a device, which is part of the appeal - but they still fail, and there is nothing to warn you first.

There is no firmware anyone updates, no configuration to back up and no credentials to manage. Over a long life that is a real saving in effort and in risk, since a device with no management interface has no management interface to be attacked through.

What does need attention is physical. Power supplies are the usual failure, particularly the external plug-top type, and cheap units in warm or dusty locations degrade faster than their specification suggests. Fanless models in enclosed positions can overheat if they are boxed in.

The practical implication is that failures are discovered by users rather than by monitoring. Where a switch feeds anything that matters, keep a spare of the same model on site - swapping one in takes minutes and requires no configuration at all, which is the other side of the same coin.

They add an uncontrolled connection point, and they hide it from whatever is monitoring the rest of the network.

Any device plugged into an unmanaged switch is on the network. There is no port security, no authentication and no way to restrict what connects. A switch added under a desk turns one controlled outlet into four uncontrolled ones.

It also breaks visibility for the upstream switch. Everything behind the unmanaged unit appears on a single upstream port, so per-device statistics, port-based access control and per-port policies all stop at that boundary.

Organisations that run 802.1X authentication find this immediately: authentication applies at the managed port, so a small switch behind it lets several unauthenticated devices share one authenticated session.

Where unmanaged switches are unavoidable, the mitigation is upstream - restrict what the port can do, limit the MAC addresses learned, and know that the switch is there. Undocumented ones are the problem.

Port count and speed, PoE budget if any, whether it is fanless, how it mounts, and the quality of the power supply.

Speed matters more than it used to. A gigabit switch in a path that is otherwise multi-gigabit becomes the limit for everything behind it.

Fanless is worth insisting on anywhere people sit or sleep, and in dusty environments where a fan pulls contamination through the chassis.

Mounting decides reliability as much as anything else. A metal-cased unit with DIN or wall mounting stays where it is put; a plastic desktop unit in a control panel ends up loose among the cables.

The external power supply is the commonest failure point, and also the commonest thing to go missing. Units with an internal supply, or with a standard barrel connector and a documented voltage, are far easier to keep running.

Finally check the operating temperature range against where it will actually be - inside a sealed panel in summer is not room temperature.