Transcript ppt

SLOW SEARCH
Jaime Teevan, Kevyn Collins-Thompson,
Ryen White, Susan Dumais and Yubin Kim
Slow Movements
Speed Focus in Search Reasonable
Not All Searches Need to Be Fast
• Long-term tasks
• Long search sessions
• Multi-session searches
• Social search
• Question asking
• Technologically limited
• Limited connectivity
• Mobile devices
• Search from space
Making Use of Additional Time
TIME + SEARCH
Understand how time influences the search experience
Time in Search: Micro Scale
• Impact of sub-second changes in response time
• Study using large-scale query log analysis
• Natural variation exists within a single query
• Measure interaction as f(response time)
• Time to click
• Abandonment rate
Small Increases  Big Changes
Relative Time To Click
1.4
1.3
1.2
1.1
Navigational
Informational
1
0.9
500
700
900
1100
Page Load Time (ms)
1300
1500
Impact of Sub-Second Changes
• Imperceptible changes impact behavior
• Impact of slower page load time
• Slower time to click
• Increased abandonment
• Some queries more sensitive to time than others
• Time impacts navigational queries > informational
• Informational queries less impacted by very slow results
1.4
1.3
1.2
1.1
1
0.9
500
1
700
10 900
1100 100
TimePage
Spent
Load
Searching
Time (ms)
(mins)
1300
1000
1500
Time in Search: Macro Scale
• Impact of longer changes in response time
• Two user surveys with 1476 participants in total
• Detailed survey with 141 MSFT employees
• Online survey with 1335 people
• Asked about the participant’s last search
• Willingness to wait for results as f(response time)
1.4
1.3
1.2
1.1
1
0.9
1
10
100
Time Spent Searching (mins)
1000
Probability of Searching that Long
Last Search Took Minutes or Hours
1.0
0.8
0.6
0.4
0.2
0.0
1
10
100
Time Spent Searching (mins)
1000
Time Constraints in Search
• Reported spending more time on some searches
• Important tasks
• Tasks with a deadline
• Tasks with bad search results
• Time constraints
• 34% of tasks needed results urgently
• 39% of tasks needed results by a deadline
• 27% of tasks needed results whenever
Willingness to Wait
• Time willing to wait < actual time spent searching
• Participants do not trust the search engine
• 28% people said they want “fast results always”
• For important tasks
• Spent more time searching than less important
• Willing to wait less than less important
Probability of Searching that Long
Willing to Wait for Quality Results
1.0
Perfect
Much Better
Better
Slightly Better
0.8
0.6
0.4
0.2
0.0
1
10
100
Wait Time (mins)
1000
Impact of Longer Intervals of Time
• People do not want to wait, but…
• Report spending a lot of time on important tasks
• 90% of the important tasks took more than 1 minute
• 23% of respondents found their results unacceptable
• Are willing wait in some cases
• When the results obtained were poor
• When a perfect answer is sought
Summary
QUESTIONS?
Jaime Teevan [email protected]
Yubin Kim [email protected]