This deck analyses user behaviour in terms of search and was originally presented at BightonSEO on Friday, 18th September 2015, written and presented by Gerald Murphy.
Topics included and discussed:
Average number of keywords typed
Search query types and user behaviour (e.g. time spend on SERP)
Search engine click bias
Why heatmaps suck and why scanpaths rock my world
Attractive, clickable keywords for metadata
Why aggregated, universal, blended, SERPs exist
Presentation of results and scanning behaviour
Snippet length and information processing
Search behaviours (e.g. pearl growth)
Male vs female search behaviours
Age
Reading time
Mobile, the environment and search
5. Types of search queries
Informational (find)
48% to 70% (highest of 80% has been researched) of all queries
Less time spent on SERP, compared to navigational queries
More document time
Navigational (get)
14% to 20% of all queries
More time spent on SERP, compared to informational queries
Less document time
Transactional (buy)
20% to 30% of all queries
Connectivity (calculate)
Newest type of query, not queried (yet !)
Google Knowledge Graph uses the theory behind this query type
@GeraldSearch
6. [bbc] example of a navigational query
Purposely non-capitalisated
7. Types of search queries
Informational (find)
48% to 70% (highest of 80% has been researched) of all queries
Less time spent on SERP, compared to navigational queries
More document time
Navigational (get)
14% to 20% of all queries
More time spent on SERP, compared to informational queries
Less document time
Transactional (buy)
20% to 30% of all queries
Connectivity (calculate)
Newest type of query, not queried (yet !)
Google Knowledge Graph uses the theory behind this query type (entities)
@GeraldSearch
10. Trust bias
Highly ranked results are trusted
Even if these abstracts are less relevant than other
abstracts
@GeraldSearchPhoto: Copyright granted, reused, unmodified
11. Quality bias
Searcher click decision is influenced by:
Relevance of clicked link
Overall quality of the other abstracts in the ranking
@GeraldSearch
15. Scanpaths
1
2
3
45 6
7
8
9
16 scanpaths
5.8 compressed scanpaths
3.2 minimal scanpaths
We don’t always view in order
of what the engine ranks
We click on what order the
engine ranks
Trust and quality bias
@GeraldSearch
24. Income and educational level
Reflective of how we engage with technologies
More money, bigger choice of technology to choose from
But have no impact on search engine use or
behaviour
Gender differences are mirrored offline and online
@GeraldSearch
25. Male searchers
Tend to spend more time examining SERP
5.4 times more likely than females to inspect lower
ranked results
More linear
View on average 3 more pages than females
More time to enter queries
Scan and filter results on SERP
26. Female searchers
Do not scroll as much
Less linear
Open 2 more browser tabs for more complex
searches
Tend to repeat views of old results
Fixate heavily on positions #2 and #3
Prefer to read target results on paper
Less time on SERPs
Browse websites more deeply
27. Age
Younger and middle-aged searchers are very similar
18 to 60s
Older spend double the amount of time on the SERP
On average an extra 4 seconds
@GeraldSearchAge development photo: Copyright granted, reused, unmodified (TheIRHistory)
28. Reading time
Indicative of interest for news stories
Reading time and scrolling equals relevance
browsing
For web information retrieval reading time is not indicative
of document relevancy because reading time differs
between subject and task
Difficult to interpret
@GeraldSearchBook photo: Copyright granted, reused, unmodified (Simon Cocks)
29. Mobile search
Good abandonment is higher on mobile search
Where searchers do not click but are satisfied with results
30% reduction in performance tapping buttons when
walking
When we walk our arms move vertically
This is why voice search exists
@GeraldSearch
30. Mobile search
Our search behaviour changes on public transport
Stronger observation on public transport
Momentarily passes on the street
Not an issue
Continue walking
Unwanted attention quickly passes
@GeraldSearch
31. Mobile
Engines will soon process mobile queries like PPC,
whereby:
Location
Time of day
Day of week
Weather conditions
Current activity of user
Temporal patterns (i.e. weekday vs weekend)
Will be factored into a mobile search
@GeraldSearch
32. Mobile in social situations
70% of searches are conducted at home or at work
Mobile search is a social activity
Often conducted in the presence of others
Device type impacts mobile search
High-end smartphones have similar patterns to desktop
searchers
@GeraldSearch
33. Key references
• Ashkan, A. and Clarke, C.L.A. (2012) Modeling Browsing Behaviour for Click Analysis in Sponsored Search. CIKM
• Belkin, N. J. (2000) Helping People Find What They Don't Know. The Human Element 43(8)
• Buscher, G., White, R.W., Dumais, S.T. and Huang, J. (2012)Large-Scale Analysis of Individual and Task Differences in Search Result Page
Examination Strategies. WSDM
• Cole, M.J., Gwizdka, J., Liu, C., Bierig, R., Belkin, N.J., and Zhang, X. (2011) Task and user effects on reading patterns in information
search. 23 23(2011)
• Cutrell, E. and Guan, Z. (no date) Eye tracking in MSN Search: Investigating snippet length, target position and task types. --
• Hochstotter, N., and Lewandowski, D. (2009) What users see -- Structures in the search engine results page. Information Sciences
179(2009)
• Practical Ecommerce (online) Why is Metadata important?
• Lewandowski, D. (2008) The retrieval effectiveness of web search engines: considering results descriptions. Effectiveness of web search
engines-
• Nettleton, D.F. Gonzales-Caro, C. (2012) Successful Web Searches: What Makes the Difference? An Eye-Tracking Study. --
• Rafiei, D., Bharat, K., and Shukla, A. (2010) Diversifying Web Search Results. WWW Full Paper
• Rele, R.S. and Duchowski, A.T. (no date) Using eye tracking to evaluate alternative search results interfaces. --
• Singer, G., Norbisrath, U., Lewandowski, D. (no date) Impact of Gender and Age on performing Search Tasks Online. --
• Sushmita, S. Joho, H. and Lalmas, M. (no date) A Task-Based Evaluation of an Aggregated Search Interface. --
• Wilson, T.D. (2000) Human Information Behaviour. Special Issue on Information Science Research 3(2)
• Gerald Murphy Search (online)
• Gerald Murphy LinkedIn (online)
Visited links increase the fixations on SERPhttp://ix.cs.uoregon.edu/~hornof/downloads/CHI04_Link.pdf
Why use a search engine in the first instance? Goal, problem
Why use a search engine in the first instance? Goal, problem
Detail, not just a snapshot
Click-through rates are affected by:
Relevance
Position of document on the SERP
Diversity of user queries
Few searchers click on facets
Too much effort
An extension of metasearch as it provides information from different sources
Broker receives query, then processes it
Relevant documents are selected for that query
Top results of each collection are merged into a single list
Images are processed first
Tubular = faster scanning
Tubular = tendency to move within columns, rather than between columns
Higher mean fixation for list interfaces, suggesting higher cognition effort
Design interfaces, make no difference to: CTRs or task type
Too much information causes cognitive overload
And irrelevant information causes cognitive noise
We do not consciously place different weight on different SERP elements
As more information is placed on the SERP, users simply subconsciously down-weight the relevance of URLs for decisionsLonger descriptions improve informational queries but degrade for navigational queries
Titles receive almost consistent views – titles are very important for SEO
Shorter descriptions place more fixation time on the URL
Clicks are not a ranking factor, we are too complex
Full of different behaviours
Males tends to look at lower ranked results, less trust and quality bias
Males are more linear and view more pages than females
Females do not tend to scroll much
Place more trust on position 2 and 3
Have more tabs open (more tabs also reduce concentration levels)
Energic walkers can have false positives – autosuggestion
Autosuggestion is there to carry out a better search
Recommendations are there to encourage use to make use of the engine’s resources
Conversations significantly impact query type
Mobile, compared to desktop, searches are:
Shorter
Of adult content
Few queries per session
Conversations prompt 27% of information seeking behaviour