Replacing Google Search Appliance: The search for better search Part 1

Google made a bold decision to stop supporting its on premise search appliance. For many of us using GSA, this means finding a replacement search solution well in advance of GSA end of life. My personal clock is ticking to expire come this September. Some of you may be sooner or later, but if using GSA you will need to address this problem.

My first step in finding a replacement was reviewing Gartner reports and starting some personal conversations with the Sitecore Community over what they use and how they like it. From this, I identified 6 potential search solution replacements.

The search solutions:
The next step was to identify criteria for how the search solution would be scored. I broke it down into a few criteria areas:
  • Search and Relevancy
    • Ranking and boosting: Ability to boost and modify the ranking of documents.
    • Query suggestion: Ability to autocomplete and suggest the term as the user types
    • Recommended Content: Ability to surface recommended content based on what other users have done
    • Query Correction: Ability to detect and correct misspelled words. Also make suggestions: Did you mean
    • Filtering and Facets: Ability to filter out search results by facets
    • Advanced Search/Query Parameters: Ability to use search prefixes and operators to refine a search – exact match, Boolean, wildcard as a few examples.
  • User Interface
    • Out of the Box Sitecore Rendering Components: The foundation is provided so that you don’t have to reinvent the wheel. Things like Search Page and various components such as search box that provides suggestions, summary of selected facets, paging or infinite scrolling results, list of facets, sort links, etc.
    • Result Templating: Support for various JavaScript template engines such as Underscore, Handlebars, etc.
  • Technology
    • Machine Learning: Ability to analyze and adjust results based on previously successful user navigation patterns. No manual tuning needed as the machine learning will optimize.
    • Usage Analytics: Capturing and reporting on search statistics
    • Cloud vs On Premise: Where the solution will live: cloud or on premise
    • Sitecore xDb integration: Ability to personalize results based on the xDB data
    • Connectors and Crawlers: Availability of connectors and crawlers to crawl the content. Main connectors needed: Sitecore, Web, File, Database
    • Full HTML Rendered Indexing: Ability to index the page as rendered to the end user. With Sitecore a lot of engines simply index the item’s fields, but there is a need to index the full rendering as sometimes additional items are rendered within the page.
  • Security
    • Document-level Security: Ability to manage security at a document level
  • Costs
    • Pricing Model/Flexible Terms: How flexible are the terms and does the pricing model make sense
    • Overall Costs: The overall costs to implement the solution. This can be a combination of things: hardware, cost to use solution, development hours needed to implement
From there, I conducted a series of calls/demos to get a sense for each of the solutions. After the 1st round of calls, I eliminated Sinequa and Attivio based on the pricing (well into six figures per year). They were way more expensive and outside of the budget I had to work with.

Check back for Part 2: Reviewing solutions: Coveo, Alpha Solutions, Mindbreeze, Lucidworks.

Check back for Part 3: The decision

Comments


  1. This concept is a good way to enhance the knowledge.thanks for sharing. please keep it up machine learning online training Bangalore

    ReplyDelete

Post a Comment