1
Banner Ads Hinder Visual Search and Are Forgotten
Moira Burke1, Nicholas Gorman2,
Erik Nilsen2, and Anthony Hornof1
1University of Oregon and 2Lewis & Clark College
2
Banners still popular, getting larger
3
Banners still popular, getting larger
4
Banners still popular, getting larger
5
Banners still popular, getting larger
6
Do banners impact search?
• “Banner blindness”: Only 20% of web users noticed any banners (Benway, 1998)
• "An animation that appears alongside primary content will disrupt your readers' concentration and keep them from the objective of your site." (Web Style Guide, 1999)
• Who is correct?
• We investigate visual search speed and participants’ recall of banners
7
Search taskSearch for linked news headlines. Analogous to:
8
Search task, cont’dOur search area:
9
Search task cont’d
• Two kinds searches: literal and semantic precues
10
Search task cont’d
• Two kinds searches: literal and semantic precues
blank
animated
static
• Three kinds of banners
11
• Two banner placements:
– top of search area– covering a random line in search area
Search task cont’d
• Two kinds searches: literal and semantic precues
• Three kinds of banners
12
• Two banner placements:
– top of search area– covering a random line in search area
Search task cont’d
• Two kinds searches: literal and semantic precues
• Three kinds of banners
• 2 blocks (by precue type)41 trials per block12 headline locations x 3 banner types + 5 practice
13
Literal precue
14
Literal search
15
Literal search
16
Literal search
17
Semantic precue
18
Semantic search
19
Semantic search
20
Surprise! A memory test
• Did you see this banner? Yes / No
• 60 banners40 appeared, 20 didn’tHalf animated, half static
• Told some banners were from experiment; others were not
21
Surprise! A memory test
22
Results: Search TimeAnimated and static were 7% slower than blank (literal condition)
Banner Type Mean Search Time (ms)Blank 2040Static 2169Animated 2193
Literal precue
Banner Type Mean Search Time (ms)Blank 6065Static 6210Animated 6110
Semantic precue
Significant difference p<.005
Not significant, but similar trend
23
Results: Memory
• Overall, memory quite poor (Bayles, CHI 2002)
• 20.1% hit rate (100% is perfect)
• 20.2% false alarm rate (0% is perfect)
• Difficulty of search didn’t affect recall
• Top banners remembered significantly better than randomly-placed banners
24
Results: Memory cont’d
• Signal Detection Theory to correct for guessing strategies of participants
• Transform hit and false alarm rates into single measure of memory strength: d'
• d' positive for static bannersd' n.s. different than zero for animated
• Bottom line: Animated ads harder to recall
25
Summary and future work
Who was right?
• Banner ads distract
• But people don’t remember them
• Animation makes recall even worse
• Currently analyzing eye-tracking data
• So far: People don’t look at banners, except by accident