Using tagging or some other type of categorization for support questions gives you a lot of data to plan improvements to your help content. And if you can combine the data on what customers are asking with data on the traffic to your Help Center or knowledge base you’re golden. What we are doing here is assuming that pageviews to a given Help Center article is an indicator of the question of a given user. This is by no means bulletproof, and there are a few reservations to keep in mind when using this method. But it does give better data than just sitting down and thinking long and hard.
So assuming you already have either tagging or categorization for your email and chat conversations in place, you now need to add the equivalent data for your help articles. First you need to connect your articles with the tags or categories you are using for your emails. So in an awesomely simple world, that means going through all your help articles and assigning one tag to them from the list of tags you are using on your support conversations. For now, we will handily ignore the fact that you will probably realize that you have several articles that correspond to no tag or to multiple tags. For now just pick one of the tags or create new ones as you need them.
The second thing to do is to get your hands on the site analytics data for your help articles. If that data is in Google Analytics, I would hook up my Google Analytics account to a Google spreadsheet through the Google Analytics API and a Google Spreadsheets script. You can also set up an Analytics report and go manually export once in a while. Once that data is pouring in, I cross reference articles with the tags I have assigned and then view data for each tag in the same dashboard where I show my account of support conversations.
This lets us look at which tags have low or high pageviews versus number of emails and vice versa. The basic patterns to look for are these. If a specific issue has a high number of contacts but a low number of pageviews, then you should figure out how you can get that article to those people so they have the option to solve the problem without having to talk to support. If you see high page views and high escalations, then the cause could be that the content in the article does not answer the question. If you see high escalations and no pageviews because there is no article associated with the tech, then that is a candidate for creating an article.
Like I said, this approach obviously builds on a bunch of assumptions. So the question to ask here is not whether this data is complete and bulletproof. But rather, which parts of this data do I trust the most, and how useful is that. We are really looking for directional information here where before we had zero information.
Another caveat is that making interpretations based on the combination of help article pageviews and escalations builds on the assumption that users progressed linearly from a product, to the help center, to contacting support. That is obviously not always the case. And especially not so if your users have the option to email you directly or start a chat from inside your product, before they even have the chance to see any help content.