Support/Search Engine Optimization

From MozillaWiki
Jump to: navigation, search

As part of improving the workflow and processes around SUMO, we're putting together a process for improving search results for the commonly-searched terms on SUMO. To be clear, this refers to the internal (Sphinx) search engine, not searches on Google or Yahoo! that result in SUMO articles.

The process consists of three overarching steps: Getting the search terms, figuring out what the user(s) were actually looking for and then adjusting/improving the results to match.

Getting the search terms

Right now, webtrends is tracking all searches done on SUMO. Every two weeks, we plan to look at the following sets of search terms:

  • Top search terms — these are terms that get the most searches in a given week.
  • Trending search terms — these are terms that get many more searches than expected given the previous 8 weeks of data.

Note: in the first couple rounds, we'll probably only look at top search terms until all of them have been addressed

Figuring out what users are looking for

We need to figure out what users are actually trying to find when they are searching for each term. We can do that by looking at:

  • Forum/Live Chat data. Do a search for that term on forum threads and see if there are common themes within the threads.
  • Google searches. Search for firefox+(search term) will see what kinds of things Google (with its extensive search optimization history) thinks users are looking for.
  • (Future) Metrics around what articles people select in search results per search term.
  • (Future) Metrics around what articles are voted helpful per search term.

Adjusting and improving the results

By improving the results, we can make it easier for users to find what they're looking for after searching. We can use the following techniques:

  • Add/remove search keywords from articles
  • Adding summaries to articles. While this may not significantly change the ranking, it makes it more apparent to users that a given article is the one that they want.
  • Split article subsections into new articles
  • Re-title articles
  • (Future) Manually force a given article to the top of a given search
  • (Future) Rerank based on how helpful an article has been voted.