Instructables’s Journey with Elastic
RICH CONTENT, WEAK SEARCH
Instructables spent years working with Solr, then tried using an external search provider, but employees remained frustrated by a search bar that returned very basic results for a wide range of queries. They were frustrated by a search engine that returned very generic, title-based results for broad searches and failed to return results for misspelled or overly specific queries. Moreover, the Instructables staff had spent countless hours tracking and quantifying the quality of their user-generated content by popularity, tags, and other metadata, but their search only returned very simple text matching results. Because search is the core navigational tool on the site and the website receives hundreds of thousands of daily searches, they needed a solution that would deliver better results.
They chose Elastic Site Search service because of its robust search algorithm, precise relevancy controls, and low maintenance requirements.
Immediately upon implementation, Elastic Site Search’s robust language modeling technology and relevance algorithm dramatically improved results for queries that formerly returned very few or no results at all. Instructables employees quickly noticed that Elastic Site Search automatically detected misspellings, even for words that are not in the English dictionary but are nevertheless common on Instructables.com (for example, “arduino”). In addition, the semantic analysis capabilities handled multiple word queries far better than their previous provider, ensuring that users would still see a full set of results for longer queries (for example, “diy stethoscope”).
DIGGING INTO THE METADATA
To improve their search results even further, the Instructables staff incorporated vast amounts of metadata information about each project into their search algorithm, indexing data such as project popularity, the number of steps involved with each project, page views, and much more. Instructables team members knew these pieces of information were critical for determining the best search results, but until Elastic Site Search gave them the tools to control how each of these attributes impacted their search results, they had no way of capitalizing on this wealth of information in their project database. Now, the Instructables team can tune their relevance model to match their expectations — after all, the Instructables staff knows better than anyone else what the Instructables community wants to see for a particular query.
After implementing Elastic Site Search on their main website, the Instructables team easily integrated the same search engine on their mobile applications. With the same engine powering search across their website and mobile applications, any customizations or adjustments that the Instructables staff makes in their dashboard immediately take effect across all of their customer touchpoints. Furthermore, the detailed analytics from Elastic Site Search help inform Instructables’s new content strategy. By looking at the top searches that return no results, the Instructables team can quantify the number of users interested in specific topics and strategize on how to go about creating new projects.