Switch to Algolia search engine
We currently use Swiftype as our search engine in docs.gitlab.com.
Reasons to switch
- Algolia provides a nicer interface and better results (needed by the support team https://gitlab.com/gitlab-com/gitlab-docs/issues/60)
- Highly configurable, we can hack our way in many things since most libraries are open source.
- It’s open source https://github.com/algolia/docsearch and we can run the crawler ourselves. John volunteered to set it up! :D
- A minimal Analytics dashboard is provided
Drawback: no analytics/statistics. Haven’t searched in depth, but I think no statistics are provided. That would be possible if we went with their (paid) plan https://www.algolia.com/pricing. Below is what we now get from swiftype.
Swiftype Analytics
-
Total searches -
Total autocompletes -
Most popular searches - What the users searched for the most in your engine. -
Most popular autocompletes - The partial or complete phrases users entered the most into the autocomplete for your engine. -
Most commonly clicked documents from search - The documents users most often clicked on across all search results. -
Most commonly clicked documents from autocomplete - The documents users most often clicked on across all autocomplete results. -
Searches resulting in the most clicks - The searches that resulted in the most total clicks into documents. -
Conversion tracking - track how search queries lead users to perform actions like reading an article, buying a product, etc. -
Top searches with zero results - indicate that users are looking for certain content on your site, but that content doesn't exist (or isn't getting successfully indexed). -
Top searches with zero clickthroughs - indicate that users are looking for certain content on your site, but can tell just from the search results page that you don't seem to have what they're looking for. -
Top searches with multiple clicks - Searches that result in multiple clicks can indicate that users simply want to see lots of different content related to the search term. However, they can also indicate that search results are misleading, causing users to click a result, only to back up and click a different search result instead, sometimes multiple times. Consider your site’s typical use cases when judging between these.