Measuring the success of the knowledge base

Chris Ilias

In March, I posted about using article feedback to improve knowledge base articles and the importance of making knowledge base articles easy to read; but those are specific areas that are part of a greater knowledge base goal, which is to make the process of Firefox self-help as easy as possible.

There are few sources of information to we draw from:

  • Top searches: The most common search terms in the SUMO Weekly metrics document.
  • Weekly common issues: Our Weekly Common Issues page tracks the most common support issues each week.
  • Article polls: At the bottom of each article, there are poll questions: “Did this article solve a problem you had with Firefox?“, “Was this article easy to understand?“, and “Please rate your experience with solving your problem on support.mozilla.com from 1 to 5” (For more precise data there’s the PageView Data.)
  • And of course, Article comments: There is a text field on each article for users to provide feedback about the article. When logged in as a contributor, that feedback is displayed at the bottom of the article.

Here’s how that data is utilized to measure the quality of the knowledge base, and make it better:

The top search terms are tested to find out if the first search results contain the article the user is most likely searching for.
If they don’t:

  • The correct article may need to be renamed to match the search term.
  • The top article in search results may be mistaken for a different issue; so a link to the correct article is added in the intro of the first search result. If users are being redirected to the correct article, the poll data should improve.
  • Keywords that match the search terms are added to the correct article.

For generic search terms the article comments for each result may clarify what users are asking about.

The weekly common issues page is checked for any items that need documentation in the knowledge base. If enough information is available to create documentation, the relevant articles are updated or a new article is created.

The comments in articles with the lowest understandability score are checked to get details on what is not understandable in the article, so we can assess what can be done to eliminate that confusion. Sometimes that means rewording or reformatting the article. In some cases it is a matter of adding screenshots. In other cases, it’s a matter of streamlining or purging the article to simplify it for users.

In the end, it’s about taking the data, analyzing why the data is what it is, and what we can do to improve each issue. As a result, the article poll scores should go up, and users will get answers to their questions about using Firefox. We’ve outlined these tests in a contributor page, so everyone as a community can be most affective in making the knowledge base better each week. You can post any suggestions for improvement in the Contributors forum.