|Users always speak about the "network" if they mean something beyond their own workstation.|
If they tell you about performance problems of the network, you should be careful.
Before changing anything you should find out what really causes the problems: the network itself - including all active and passive components - or only the involved computers, that might be overloaded due to various reasons.
The next question is - does the network itself meets the SLA's, your propagated service level agreements?
Imagine, if some automatic supervision of your network would have stored all relevant information of the last days or weeks. Then it would be an easy task to spool back and find the date and parameters of your networks when the customers had still been happy and satisfied. And you can find out what parameters have changed.
Having acquired a sort of black box for the network at the right time is a great thing.
If you use a switched or routed network with SNMP capable active devices, it would be nice to have a program which recognises all changes and searches the network devices automatically, ask them about their ports, and read out all important port counters (netload, packet and broadcasts etc.) and record them round-the-clock.
You can then search the collected data segment by segment and see when the problem was reported first. This will save you a lot of time.
If you have a high number of broadcast packets, you can search for their producers with a network monitor like NetControl. NetControl uses hitlists to show the stations which are responsible for the highest traffic, the most broadcasts etc. Finally a separate protocol analyzer may be necessary to find out, what kind of broadcasts are used.
If the network does not show problems, this will point to a configuration problem of the involved devices.