The influence of snippet length on user behavior in mobile web search. An experimental eye-tracking study
| Published date | 20 May 2019 |
| Pages | 325-343 |
| Date | 20 May 2019 |
| DOI | https://doi.org/10.1108/AJIM-07-2018-0182 |
| Author | Jacqueline Sachse |
| Subject Matter | Library & information science |
The influence of snippet length
on user behavior in mobile
web search
An experimental eye-tracking study
Jacqueline Sachse
Berlin School of Library and Information Science,
Humboldt-Universität zu Berlin, Berlin, Germany
Abstract
Purpose –Web search is more and more moving into mobile contexts. However, screen size of mobile
devices is limited and search engine result pages face a trade-off between offering informative snippets and
optimal use of space. One factor clearly influencing this trade-off is snippet length. The purpose of this paper
is to find out what snippet size to use in mobile web search.
Design/methodology/approach –For this purpose, an eye-tracking experiment was conducted showing
participants search interfaces with snippets of one, three or five lines on a mobile device to analyze
17 dependent variables. In total, 31 participants took part in the study. Each of the participants solved
informational and navigational tasks.
Findings –Results indicate a strong influence of page fold on scrolling behavior and attention distribution
across search results. Regardless of query type, short snippets seemto provide too little information about the
result, so that search performance and subjective measures are negatively affected. Long snippets of five lines
lead to better performance than medium snippets for navigational queries, but to worse performance for
informational queries.
Originality/value –Although space in mobile search is limited, this study shows that longer snippets
improve usability and user experience. It further emphasizes that page fold plays a stronger role in mobile
than in desktop search for attention distribution.
Keywords Information behaviour, Eye-tracking, Search engine result page, Mobile search, Snippet length,
Web search
Paper type Research paper
Introduction
Web search is more and more moving into mobile contexts and thus carried out in
environmental settings often different from desktop search. For instance, since the majority
of mobile searches are conducted while the user is on the move, i.e. away from home or office
(Komaki et al., 2012; Nylander et al., 2009; Westlund et al., 2011), it is more likely that search
happens in situations with stronger distraction as well as less time and concentration for a
careful examination of search results. Additionally, more scrolling is involved on smaller
screens (Kim et al., 2016) which can be expected to increase the cognitive workload when
results get scrolled outside the viewport. As a result, it becomes even more essential for the
success of mobile searches to provide a smooth user experience and to minimize user effort.
In terms of relevance evaluation, searchers mainly rely on a result’s snippet, and so the click
decision is usually based on only a few lines of description text next to title and URL.
However, since screen size is limited, search engine result pages (SERPs) have to face a
trade-off between offering informative snippets and an optimal use of space. One factor
clearly influencing this trade-off is snippet length. Search engines currently typically
display snippets of three to four lines. Longer snippets can provide more information about Aslib Journal of Information
Management
Vol. 71 No. 3, 2019
pp. 325-343
© Emerald PublishingLimited
2050-3806
DOI 10.1108/AJIM-07-2018-0182
Received 30 July 2018
Revised 12 February 2019
Accepted 13 February 2019
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/2050-3806.htm
This study was supported by eye square with premises, technical equipment and expertise in
eye-tracking technology.
325
The influence of
snippet length
in mobile web
search
a search result. In some cases, the desired information might already be found in the snippet,
thus rejecting the need to click on the result. At the same time, longer snippets entail that
less results fit on the screen. By contrast, shorter snippets make room for more results and
allow users to compare a larger number of results more easily. However, shorter snippets
provide less contextual information.
This study demonstrates what snippet size creates the best user experience, i.e. enables
users to find the desired information easily and quickly with the least effort. To address this
issue, an eye-tracking experiment was conducted in which 31 participants each executed ten
tasks on a smartphone[1]. Tasks were either navigational or informational, as defined by
Broder (2002), and were performed on manipulated SERPs showing snippets with a length
of either one, three or five lines.
Related work
Psychological tests indicate that eye movements strongly correlate with cognitive processes
( Just and Carpenter, 1976). Hence, eye-tracking is a powerful tool for the study of user
behavior and has been applied in many studies in this field (which are discussed below in
this section) coming with some established assumptions and terminology. The moment the
eye is relatively stationary is called a fixation (Lorigo et al., 2008; Rayner, 1978, 1998).
The minimum duration for a fixation is typically 100 ms (Granka et al., 2004; Hall-Phillips
et al., 2013; Kim et al., 2015; Poole and Ball, 2006). The research object (e.g. a website or
software) is divided into several areas of interest (AOIs). In the context of SERPs, it may be
expedient if an AOI consists of one search result. The scanpath describes the order of visited
AOIs. The length of a scanpath tells how many fixations a user made. To determine how
many results they have seen, Lorigo et al. (2006) introduced the minimal scanpath. It is a
reduced version of the scanpath without any backward jumps (returns to previously
regarded results) or fixations within the same AOI.
One of the most fundamental findings regarding general scanning behavior is that users
scan the SERP linearly from top to bottom (Aula et al., 2005; Dumais et al., 2010; Granka
et al., 2004; Klöckner et al., 2004) and tend to click immediately once a relevant result is
spotted (Craswell et al., 2008; Klöckner et al., 2004). Klöckner et al. (2004) coined the term
depth-first strategy for this mode of scanning in contrast to the breadth-first strategy, which
involves a consideration of all results before the click decision. Further, previous research
consistently showed that the position of a result within the result list is one of the strongest
determinants to gain users’attention and clicks (Granka et al., 2004; Guan and Cutrell, 2007;
Joachims et al., 2005; Klöckner et al., 2004; Pan et al., 2007). Aula et al. (2005) called looking at
top results only “economic strategy”and the examination of further results “exhaustive
strategy.”However, there are studies suggesting that the ranking bias can be weakened by
several factors, such as a bias against ads (Buscher et al., 2010; Djamasbi et al., 2013; Dumais
et al., 2010; Hall-Phillips et al., 2013; Hotchkiss et al., 2005; Jansen and Spink, 2009; Jansen
et al., 2007), the results’relevance (Dumais et al., 2010; Keane et al., 2008; Lorigo et al., 2008;
O’Brien and Keane, 2006; Schultheiß et al., 2018; Wu et al., 2014; Yue et al., 2010) or a domain
bias (Ieong et al., 2012), all of which can encourage users to skip the first few results and to
consider lower ranks as well.
SERP design for mobile devices faces limited space, touch-control and a shorter
attention span. The latter can be addressed with instant-answer features or a knowledge
graph, both of which enhance user experience (Lagun et al., 2014; Navalpakkam et al.,
2013). Limited space raises the question about a suitable snippet length. A look at desktop
search might be helpful: While some research indicates that longer snippets can be
beneficial for desktop search (Dumais et al., 2001; Jones et al., 2002, 2003; Paek et al., 2004),
there is one particular study pointing to a more complex situation. Cutrell and Guan (2007)
experimented with short, medium and long snippets on desktop PCs. They found
326
AJIM
71,3
To continue reading
Request your trial