Improving site search with search analytics
8 likes2,342 views
This document discusses improving site search through search analytics. It begins by outlining the basic steps to be as good as Google for site search: index content, design a search form, and show the best results first. It then discusses analyzing search query logs from a site to understand user behavior and improve search performance, content, navigation, and user experience. Specific techniques recommended include analyzing the most common queries, randomly selecting "long tail" queries, measuring relevance of results, removing irrelevant results from the index, using metadata and synonyms, adjusting the ranking algorithm, and boosting or reducing the score of individual results.
Improving site search with search analytics
- 1. Improving site search with search analytics Audun Rundberg June 22nd, 2013
- 6. Be just as good as Google - in 3 small steps: 1. Index everything you want to search 2. Design a search form: 3. Show the best results first when people search on your site Search!
- 10. Photo: Google
- 12. B 38,4% C 34,3% F 3,9% E 8,1% D 3,9% A 3,3% 1,6% 1,6% 1,6% 1,6% 1,6% Source: http://en.wikipedia.org/wiki/PageRank
- 13. Content Links Domain history User behaviour Social media sharing
- 15. Google vs You Google • Billions of documents • User behaviour data from hundreds of millions of users • 50.000+ employees Your search engine • 100-100.000 documents • Probably not used very much • 0-1 employees
- 18. #1 Check your search results
- 19. your site, they’re telling you in their own words m your organization. In Search Analytics for Your Site, ws you how to harvest and analyze search query ur site’s search performance, content, navigation, ness, and overall user experience. dback will improve any site’s user experience, why aren’t we spending ords our audience uses when asking us for stuff? I can’t imagine a more Rosenfeld to help us put this amazing data to work.” te lays out pretty much everything you need to know to mine your query sitive customer experience.” Demystified lytics for Your Site is a superlative work from the initial story to the final lytics and UX practice. Great book!” oks that makes me pound the table with my fist and yell, ‘Yes! Exactly! g it.” nt Strategy for the Web OR YOUR SITE ks/searchanalytics/ SE ARCH ANALY TICS FOR YOUR SITE Conversations with Your Customers by LOUIS ROSENFELD forewords by Steve Krug and Avinash Kaushik SEARCHANALYTICSFORYOURSITEbyLOUISROSENFELD
- 21. 0 62,5 125 187,5 250 1 9 17 25 33 41 49 57 65 73 81 89 97 105 113 121 129 137 145 153 161 169 177 185 193 201 209 217 225 233 241 249 Top 50 queries: 10% of searches 50 of 23,252 total queries: 0,2%
- 23. PPM
- 24. PPM 5. place... Good enough?
- 25. Query Best result vacancies 1 sia 2 ppm 4 plane tickets 24 travel 3 Average 6,8
- 26. Old – Jan 31 New – Jan 31 New – Feb 22 New – Apr 17 0 2,5 5 7,5 10 4,2 9,3 1,74 1,27 Average position of best result
- 27. Most common queries 1. Find the 50 most used queries in your search statistics, or as many as you have time for 2. Perform a search for each query 3. Decide which result is the best 4. Write down the position of the best result 5. Calculate the average 6. Repeat from time to time
- 31. Take 100 randomly selected queries
- 32. «Long tail» queries 1. Take 100 randomly selected queries from the «long tail» 2. Categorise them 3. Perform a search for each query 4. Decide which result is the best 5. Write down the position of the best result 6. Calculate the average 7. Repeat from time to time
- 33. Relevance
- 34. How to measure 1. Find the 50 most used queries in your search statistics, or as many as you have time for 2. Perform a search for each query 3. Rate the five first results from A to C A. Perfect result B. Not perfect, but the user will get to the right page C. Irrelevant
- 35. Relevance Query Result #1 Result #2 Result #3 Result #4 Result #5 Relevance (strict) Relevance (less strict) mesys A B B B B 20 % 100 % glasses A B C C C 20 % 40 % cell phone A C B B B 20 % 80 % Average 20 % 73 %
- 36. #2 Improve your search results
- 39. «hytte» «company cabin» Result 1 - 66.762 points: Date: 923 points IA: 60.000 points Word matches: 66 points Title matches etc.: 5773 points Result 2 - 62.637 points: Date: 946 points IA: 56.000 points Word matches: 31 points Title matches etc.: 5660 points
- 40. Remove irrelevant results • Front page • Admin pages • Different views of the same page (e.g. screen and print versions) • Duplicates
- 41. Metadata matches • Results that match the user profile property «Location» get a higher score • Also good for filters
- 42. Rewrite content • «training» instead of «competence-increasing measures»
- 43. Use synonyms • Create a company-specific list of synonyms that your search engine doesn’t already know about
- 44. Adjust the algorithm • Title match is more important than a text match
- 45. Boost individual results • Quick solution if a common search query gives bad results • You need a plan for how to review manual changes
- 46. Give unimportant content a negative boost • Older news articles or archived content receives a negative boost
- 47. Buy a new and more expensive search engine that will solve all your problems Let me know if you find one!
- 48. Thank you!