Site Search Considerations – in a Sitecore XP and Solr context

Alpha Solutions primarily relies on search in the solutions built. We also have a large demand for consulting, specifically with Sitecore XP and Apache Solr combined. The main goal in this post is to share thoughts and considerations that you, as the client, could benefit by taking into account while establishing a Site Search while also bringing your current search to the next level. Hence, this will primarily be a non-technical post.

What is your search used for?

Search has changed the way users expect to find what they are looking for. How often do users really click down through navigation hierarchies? As users, we expect find the right results on the first search; or as Google put it: “I’m Feeling Lucky.”

When looking at your own site(s); what is your search used for? Is it simply a search feature that users can choose to use, or is it central to your site?

Alpha Solutions believes that a strong yet flexible search, with combined personalization along with use of other strong features of Sitecore XP, can be a game changer on most sites.  Whether it is content focused, ecommerce and quite possibly both.

As a business owner, your goal is to get your users to the right content fast and easy. Technology will not stand alone to enable that, but the right choices can help in the near future as well as the distant future.

Where do I start?

The first step in the process would be the analyzation of your own site(s), gather as many facts and data possible. That alone will show you a lot about the current usage, and also aid in obtaining comparable data that can be used at a later time.

In most cases we complete the following tasks:

  • Investigate current search analytics, along with the web analytics numbers to investigate how each search is being used
  • If search is not tagged appropriately, we advise on what should/should not be tagged to enable collecting said search analytics
  • Review Sitecore analytics for even more useful info
  • Use IIS logs and/or logs from possible current Solr installations and even other search engines currently in use

Following the implementation of previous tasks, we evaluate the current search UI:

  • Is the search field placed prominently, is the field large enough?
  • Is auto suggest being used?
  • What information is being shown on after each search result?
    • Is that information relevant to the given context?
    • For ecommerce: The purchase of buttons and images is key.  If applicable to said site, are they easy to use or navigate?
  • Filters and facets:
    • Is the placement good?
    • Is it easy to selected/not selected values?
    • Is the order of the facets relevant?
    • Is it the use of relevant filters/facets that bring users forward to find the wanted results?
    • Could the filters or facets be dependent on the users’ profile or possibly persona?
  • Is sorting and ranking available for use?
  • Are results relevant when sorting and ranking, within the given context?
  • Has spellcheck been used?

Refer to the above questions to help kick-start a business-oriented discussion. Be aware that the level of knowledge around can vary.  Which is not a problem at all, it is in your favor to take the time needed to spend with your business owners to educate their basic search understanding, if need be. Obtaining this kind of information can prove to be key in understanding the business as well as their expectations/goals for you.

What content do you have available for search?

Content is as always “King” when it comes to delivering good search experiences. Another major step we use is to look at what content we have to work with.  How is it structured? To aid in obtaining pertinent information stated above; refer to below questions

  • What content is available for the search engine?
  • Is the content provided in your results enough? Is it able to be used in your favor to further detail like brand, designer, category descriptions, etc.?
  • Is content and searched structured, i.e., (not as one big, oversized “text,” or “content” field); instead of matched, relevant, along with individual fields?
  • Are the available languages us for content translation relevant?
  • Did you search inside documents used? (i.e., PDF, word, etc.)

**Important things to consider:**

  • Specific search content used, does not need to be shared with users….. However, such search specific content can be crucial in helping search engines locate the most relevant results for users. This “enriched content”, can positively influence in ranking of your search results. For Example, perhaps you run an ecommerce site and have pertinent information regarding storage cost, along with margins gained from each product sold. Collectively, these things can be utilized in your favor.  Another consideration that is useful   could simply be the use of information found from the user’s prior visits, content looked at, and even detailed product bought.

Search Relevance Tuning

Search relevance, ranking & tuning – these terms are essential to reaching the end goal: To make the users find the needed content with as little effort as possible.

We want the search engine to provide the best results possible to users, based on the users’ query. Search Relevancy tuning is the core of that, hence a major factor that plays a role in having a lot of questions. See list below of some examples of this:

  • What are the current issues? Is the recall too low/high? How relevant are the given results?
  • What data is available for ranking? i.e.: price, earnings, content, promotional info (is the product part of a promotion), a specific brand, is each usage description present?
  • How is ranking used/tuned?
  • How quick are the search results given?
  • Are important products/categories of specific products boosted?
  • Do you use price/earnings as factor in boosting?
  • Do you use freshness? How new are the listed products?
  • Do you differ the rank according to user persona?
  • Do you tune the top 100/250 search terms used?
  • Things like stemming, spelling correction, and proximity might affect recall, but also give noise
  • Do you use synonyms?
  • Do you protect proper names from stemming, etc.?
  • Search relevancy tuning exercise should be run iteratively (month, month, 2 months, 2 months and the quarterly from there). When did you last do it?
  • Document both good and bad search cases. Base these on experience and search statistics, as well as inside knowledge

Use the above questions to get as much information about the current search implementation. Also, use it as a setoff to talk about what is possible to do with search. Spend time to define goals, and a search strategy. Remember to set goals that can be measured. Finally, start the process of creating a roadmap to begin implementation. Evaluate effort, cost, and gain for each initiative.  These minor details help in gaining process and use the data to make it measurable.

NOTE: Search ranking and tuning is primarily about analyzing, and trying different configurations. Make sure that you know your test index. In most solutions, the search index changes all the time, even during updates. In that case it’s hard to compare the various changes you implement over time. Track a number of test indexes for testing, development, and evaluation.

Search monitoring and operations

Finally, touch on how you are planning to monitor the searches requested. There’s a lot of technical details in monitoring that is valuable for business growth.

See below for just a couple of examples:

  • Alpha-Solutions often builds solutions that rely heavily on search. That means, that to show a single page, we might have to test a lot of searches (e.g. list, latest news). The queries that hit the search engine from those are often irrelevant. We need to follow the searched details that the user initiates from typing in a word, phrase, or filtering etc. We distinguish the various search as “technical searches” and others as “user searches”.
  • Queries trigger from spiders, crawlers and surveillance tools – these are all technical searches. You need to be able to filter those from your analytics.

To get more information on the monitoring that the client currently has with a current solution, we initiate the discussion with questions much like the ones listed here:

  • How do you monitor?
  • Do you distinguish between technical searches and user searches?
  • Do you test the changes you do to your search functionality?

What about:

  • Ranking
  • New order of facets
  • New data in search index (and therefore in search result)
  • Are you satisfied on the current monitoring? Do you use the data you get?
  • What is the reason for the “success” of a given change?
    • The buy conversion?
    • The click-through to a search result?
    • The further use of search features (using a new facet, because it became more relevant)?

Define what and how to monitor the search. Be clear on the areas where adding logic to the solution is needed to get relevant analytics.

Summary

In this blog post we have shared some ideas and specific questions that we use as a part of discovery/clarification phase on site searches. We are always interested in comments, suggestions, and experiences that you have had. We ask and welcome you to please add your comments!

 

Roland Villemoes & Klaus Petersen

Alpha Solutions

Site Search Considerations – in a Sitecore XP and Solr context

Leave a Reply

Your email address will not be published. Required fields are marked *