Back to top

Learning from E-Commerce: Search Quality Improvement

Online commerce is booming – what is the impact of search and what can search practitioners in other fields learn from this?



  • Search provides a clear and measurable ROI in e-commerce
  • Analysis of e-commerce search behavior provides valuable direction for other search applications



The value of search in e-commerce is obvious.  Most industry studies will confirm that:

  • Almost half of online shoppers use site search while shopping online
  • Visitors who use search are more likely to make a purchase
  • A bad search experience almost guarantees abandonment – a lost sale
  • A good search experience increases conversions and average order value


So the ROI of good e-commerce search is clear and measurable. But what about the value of search in other applications? What is the value of search within the enterprise? Based on our experience with over 500 search clients, the goal of enhanced search is often improved productivity, as illustrated by the following examples:

  • Employees who waste much of the working day hunting for information across multiple silos
  • Researchers who repeat experiments already conducted
  • Sales staff who don’t have access to the best proposal materials, including past similar proposals
  • Recruiters who have to navigate multiple search interfaces to match candidates to jobs


When compared to e-commerce, it may be difficult to always quantify the exact ROI of improved search in these use cases. However, we can learn something from the techniques used for search improvements in e-commerce, where there is a “measure everything” mentality.  

Every interaction with a website produces a log file entry, and the corpus of data built over time can be analysed in numerous ways. It is an ideal environment for measuring the difference that an improved search capability makes. 



In seeking to improve search systems, e-commerce folks often start by examining search logs, and categorizing queries. Appropriate strategies can then be used to improve search relevancy for each category in turn. This is a valid approach for any substantive search system, even for corporate-wide / intranet systems serving a diverse range of user needs. 

The overall methodology is simple:

  • Capture and process search log files – the more the better, although ideally the data should be relatively fresh
  • Working with people who understand the domain and subject matter, categorize queries into a number of “query types”
  • Create and implement a search improvement strategy for each query type, starting with the most numerous, and then working down the list

This approach helps to bring some focus to search improvement. “Poor relevancy,” which is the most common complaint made against search systems, becomes easier to address through query categorization. A generic request to “fix the relevancy” can otherwise seem daunting.

To illustrate the approach, here is a brief description of three typical e-commerce query types, and a summary of the strategies that might be employed to improve relevancy.


Examples:  Dell Latitude E6530,   or  Fender FSR Aerodyne Stratocaster

For this query type, the user knows exactly what they want, and expects to be taken directly to the appropriate product. Measures to ensure success (assuming that you stock the product) can include:

  • Making sure that full catalogue details are indexed and searchable, but in a balanced way. Some data is more important than other data
  • Catering for common abbreviations and synonyms
  • Using query auto-completion and/or automated spell checking to help the user avoid unintended typos
  • Pre-processing queries to remove extraneous characters caused by cut and paste (often, exact product searches are cut from another web page)


Examples:  laptop computer,  or  red dress

In this situation, the ideal search response will show a variety of options to the user, and also provide navigation tools to help the user explore the available products. Tactics can include:

  • Working hard at capturing or creating metadata to populate navigation options and provide alternative sorting capabilities
  • Providing results in an easily browsable format, with appropriate use of graphics (depends on the product / service)
  • Enhancing the query with applicable alternative descriptions that mean the same thing, to ensure full recall. For example Laptop PC, or scarlet dress


Examples:   Ruggedized Laptop PC,  or  red cotton dress size 10

e-commerce queries will often combine an exact or generic product search with one or more qualifiers. Advanced query parsing (using tools such as QPL) can help recognize and interpret these qualifiers, or “features”, and where appropriate, map them to specific fields for search purposes. 



You probably already have the technology in place to create search excellence – all of today’s leading search products are highly capable and most are feature rich. Applying some thought and effort to tuning your search system can be transformative for productivity and user satisfaction. Regardless of your search application type and objectives, we can all learn from the e-commerce folks.

What are the search categories that your system has to cope with? Logfile analysis is the place to start, and if you need any help with how to approach this, or which tools to use, contact us about our search quality analysis service.