We often speak of knowledge management in terms of keeping a usable knowledge base for support or making customer information more useful than basic demographics. But knowledge management encompasses more than just arranging information to be used for customer service, either as self service or for agents to use while handling issues over the phone or chat lines. It is even more than a way for marketers to segment the customer base.

If you aren't running a report that gives a mathematical and graphical representation of a trend, all you have is anecdotal evidence from customer support.

Important information can be gained by keeping track of the reasons for customer complaints.  Develop a classification system and start separating problems into buckets. With this information you can start analyzing complaint trends that come through the customer support center.

For example, if you support a machine that consists of many individual parts. You may record the machine, the part name/number that failed, and how it failed:

chemisty analyzer > probe > lot number > bent out of box

Why keep this level of detail for every call? Doesn't sound like a real problem. And it may not be if that is the only call about a bent probe of that lot number for that chemistry analyzer. But what if 50 customers call in one day to report the same thing?  Or 50 customers over the next 30 days? Would you be able to pick up on that? And if you did see a spike in calls would you be able to trace it to one particular problem?

If you aren't running a report that gives a mathematical and graphical representation of the numbers, all you would have is anecdotal evidence from customer support,"Wow, I sure have answered a lot of calls about bent probes this week." But no justification for investigating. The probe would get replaced for free and customer support would continue to get calls and replace probes.

But maybe a systemic issue is occurring. How many probes will you continue to replace before noticing? How much will that cost over time? And how will all this come to light in a way that allows the issue to be observed and resolved?

Code or enter each bent probe into a database system that allows a report to be pulled using that failure code against that part over time. You will see not only that a trend is occuring, but when it began, if it is still occurring, and if it affects more than one lot, part, or machine. The lot number or date code can help narrow it down to a particular manufacturing or packaging line and streamline the process of identifying the problem area.

Now you have the information you need to investigate before the problem becomes worse. You can track down what the failures have in common and pinpoint the area where the damage is occuring. Get that fixed, make certain it can't happen again, and the part failure should stop. Next month you run another report and see that the trend has decreased.

Now you have evidence both that the trend occurred. You identified the problem, fixed it, and have the numbers to show how much money was saved in the process.