Clicking through overload: When choice overload can actually increase choice
- First Online:
- 190 Downloads
The existence of choice overload in the behaviour of online consumers is investigated through an experimental product category (electronic pencil sharpeners) on a fully functional online retail website for office supplies that was created for the field experiment. Emails were sent to potential customers with an assortment of product descriptions and links that led to a product category landing page on the retail website where visitors were then presented with an assortment of products. We find that the likelihood of click-through (either in the email or on the landing page) at first increases with the number of choices presented and then decreases consistent with choice overload. However, we also find that click-through subsequently increases as the number of choices increases after overload occurs. This suggests that another effect is also at work in the choice situations. We provide an explanation and evidence for this post-decline increase in click-through — specifically, as the number of link choices increases, not only does choice overload increase, but also the promise of richer information, which can help resolve or reduce the overload for the mere ‘low cost’ of a click.
Keywordschoice overload information overload online consumer behaviour click-through behaviour online retailing
The number of books to choose from in a typical library is both wonderful and terrible. Wonderful because of the vast amount of information available to read and terrible because the sheer number of choices can be overwhelming — so overwhelming, in fact, that Melvil Dewey in 1876 developed a classification system to organize it all for the purpose of making it more accessible. It is no wonder that the condition of ‘library anxiety’ causes a sense of powerlessness when beginning an information search.1
Similarly, shoppers appear to face what some researchers have deemed ‘information overload’.2 Loosely put, ‘information overload’ describes the state an individual reaches when the amount of information available exceeds his or her ability to process that information within an allotted time. Information overload was originally calculated as the number of options within an assortment multiplied by the number of product attributes — choice overload is a special case of information overload where only the assortment is large.3
This paper presents a study where we observe actual online consumers experiencing choice overload. We avoid a potential price-demand endogeneity issue5 by exhibiting full control over pricing in our own online store instead of observing the price choices of an independent retailer. The results of this study show that the likelihood of click-through in an email or website initially increases as the number of links to click increases and then decreases as the number of links presented continues to increase, consistent with choice overload. However, they also show a surprising subsequent increase in click-through as the number of alternatives offered increases beyond the point of overload. Next, we discuss the definitions and debate regarding information and choice overload.
Many decades ago, Miller6 provided evidence that humans can process about seven pieces of information, plus or minus two. Later research has more rigorously addressed this issue and suggested that that number is closer to four, plus or minus zero.7 When the information available to process exceeds human ability to do so in the time allotted, this can lead either to sub-optimal choices or to coping strategies such as delaying choice.
Ironically, evidence presented by Malhotra,18 which used the same basic design as Jacoby et al., found that when consumers were asked to choose between houses, choice accuracy decreased when the number of attributes to be evaluated increased to 15 or more. Muller,19 on the other hand, reports the results of a two-week field study of point-of-purchase signs hung in the aisles of grocery stores and finds that varying the amount of information on these signs produces no significant change in purchase behaviour. Lastly, Keller and Staelin20 demonstrate that providing more information about product attributes can both help and hinder choice accuracy when they break information into quantity and quality. According to their findings, quantity of information decreases choice accuracy while quality of information can increase choice accuracy.
More recently, Timmermans21 demonstrates that when choosing among multiple alternatives, subjects were more likely to use an elimination strategy as the number of alternatives increased. Hauser and Wernerfelt22 suggest that as both the number of alternatives and the information about those alternatives increases, consumers tend to consider fewer choices and to process a smaller fraction of the overall information available. Additionally, research on choice overload23 demonstrates that as choice-making complexity increases with the number of alternatives in the choice-set, individuals are more likely to avoid making a choice. Iyengar and Lepper23 also note, after citing numerous studies that demonstrate only a positive correlation between the number of alternatives and the likelihood of choice, that ‘the number of options presented in previous [choice] experiments was characteristically small, typically between two and six alternatives. It would appear, then, that what prior research has actually shown is that relatively limited choice among alternatives is more beneficial than no choice at all’.
The existence of choice overload has been supported in a retail setting, but do these effects exist in an online environment when the time and cost of click-through is considerably less? Haynes26 found evidence for choice overload when he constrained the decision makers’ time to make a decision. Most often, online shopping is an activity that happens at home with relatively fewer time constraints compared to a typical retail environment and, as such, it is worth examining whether overload exists in the electronic environment under this assumption.
Bricks and mortar retailers have a physical limit on the maximum number of products to display, but online shopping sites have virtually no limit.27 One of the advantages of the internet retailer is the ability to convey large assortments and much information at lower costs, which can reduce the cost and effort for consumers to search.28 As a result, researching the behaviour of online shoppers and their reactions to varying quantities of information and choice is relevant for today’s e-market.
Our study of information and choice overload is more related to the body of internet research that studies ‘within-site search’ behaviour. Other studies related to our research also fall within this category. Ansari and Mela30 examine the effect of link order and placement on the probability of clicking a link. Clicking a link is a ‘hit’, as defined by Berthon et al.,31 where surfers land on a site but do not necessarily interact with the information or options available. Nevertheless, the longer they stay there, the higher the possibility that they will move on to explore the site.32
Lee and Lee37 believe that this ability to access seemingly endless amounts of information has compromised the utility of online information for many consumers and may result in overload. Yet, the cost of choosing to click a link is low in terms of time and effort, and includes the ability to reverse the action quickly and easily with the back button. Because of this, the propensity to choose may increase as overload increases due to the low click-through costs. Our research fits here by examining the effect of the number of choice alternatives offered while browsing or searching a site (ie number of links offered) on the likelihood of continuation (ie clicking a link).
The framework underlying our field experiment is simple. If individuals experience choice overload, we should be able to create a situation where, when they are presented with a choice situation, they will be less likely to choose one of the available alternatives due to this overload and, thus, will be more likely to exit the choice situation. In order to accomplish this, we need to provide choice alternatives in an environment where we can vary the quantity of information associated with the choice situation, observe choices and show that, as the information provided increases, individuals reach a point where choosing one of the available alternatives becomes less likely.
To be more clear, if the probability of clicking a link increases initially, but then reaches a peak and thereafter begins to decrease (all while the number of links continues to increase) even when controlling for other factors such as link placement and order, such a reverse in the effect should indicate that individuals are experiencing information overload.
However, individuals do not have unlimited mental capacity and thus there should come a point as the number of alternatives increases above some optimal number, say n*, where considering n*+1 alternatives involves mentally processing more information than the individual is capable of processing in the time he or she is able or willing to devote to the choice. At this point where the choice becomes more complex, but the resources used to make the choice cannot be increased, we should observe a decrease in the probability of making a selection instead of leaving the choice situation.
The simple explanation for this decrease in the probability of choosing an alternative (instead of exiting the situation) is that, once an individual reaches n*+1 choices and becomes overloaded (and thus is constrained because he or she can no longer increase the amount of mental resources devoted to the choice at hand), his or her probability of choosing any available alternative cannot be greater than the probability of choosing an alternative when the same number of mental resources is distributed over n* choices (ie the unconstrained case).
The data for this research were collected through a field experiment run on a fully functioning internet retail website that was created solely for the purpose of this experiment. The online retailer sold office supply products and was designed and run by a firm that specializes in creating such websites. This ensured that the website was of professional quality and had all of the characteristics of a typical online retailer (eg menus, product pages, shopping cart, checkout process, credit card payment, a home page with promotions, customer accounts).
most people are not familiar with the key brands in the category and therefore this reduced brand salience as a decision-driving attribute;
we could offer competitive prices in this category; and
we could offer many choices without changing the key benefit driving purchase (ie sharpening pencils).
Emails that promoted the sharpeners were sent to 3.32 million addresses from a purchased list of US-based email addresses. The emails contained only text and for each sharpener they included the brand and model name, a brief description, a price and a link to the sharpener menu page on the website. Four types of email were sent that varied by the number of products they advertised. The emails contained two, four, seven or ten sharpeners, and each email type was sent to a fourth of the address list. Within each email type, two different emails were used that randomly varied the sharpeners included and their order of appearance in the email. When respondents clicked a link, they were taken to a website landing page that displayed a menu of up to nine sharpeners. The selection of sharpeners displayed and their order were randomized for each new respondent.
Clickstreams for each visitor were recorded. To control for learning through repeated visits, the clickstream data were filtered to drop all data except first-time visitors who entered the store through one of the experimental emails. A total of 1,368 unique individuals visited the website from email links.
Analysis and results
Dependent variable: was the email clicked?
4 Email Links
7 Email Links
10 Email Links
Dependent variable: page views after entering the website at the landing page
2 Menu Items
3 Menu Items
4 Menu Items
5 Menu Items
6 Menu Items
7 Menu Items
8 Menu Items
9 Menu Items
4 Email Links
7 Email Links
10 Email Links
If this explanation were correct, we should expect to see a difference in who makes a second click (after entering the website) depending on whether they came from the two-, four-, seven- or ten-link email. Individuals who received a two-, four- or seven-link email should be less likely to bounce (ie less likely to request an additional page view once they arrive at the landing page) because they are interested in the products advertised in the email. Furthermore, they should be less likely to bounce as their interest increases or, in other words, as the number of email links increases, up to the point of overload.
Dependent variable: second click into the website (ie not bouncing)
4 Email Links
7 Email Links
10 Email Links
As stated, we suggest that this difference in co-efficients appears because many subjects entering the website from the ten-link email are entering the site for a different reason than the other subjects. Specifically, we suggest that they enter the site to obtain richer information in general (which is different from the others who are more likely to be entering out of interest in the site or a specific product), and therefore a greater portion of them bounce once their information needs have been met and they have decided that they are not interested in the products offered.
In this paper, we provide evidence of choice overload in a real-world, online retail setting. We also find an additional effect, specifically that as the number of alternatives increases beyond the point where overload begins, the likelihood of clicking a link first decreases, but then begins to increase again. We propose an explanation for this effect, specifically that this second rise in the likelihood of clicking a link results from individuals seeking richer information before making a choice. Because of this different reason for entering the website, these individuals are also more likely to bounce from the landing page.
Due to limitations placed on the researchers by the email address vendors, only four types of email could be sent out, each with two revisions. Ideally, emails would have had between one and ten links and each of these ten types of email would have had many more revisions so that emails would have been completely randomized across products, product order and number of products offered. As the data currently stand, the potential order and product effects do not seem to affect the results, since we observe a similar effect in the second stage where randomization was performed, but it is difficult to provide stronger evidence that this statement is true.
Research can and should be extended to see how information overload affects initial behaviours on the internet such as clicking a link, and also to see how information overload affects the likelihood of purchase. This was not possible in the current study due to low traffic and purchase rates on the site, and thus would require creating a site with much more traffic to be successful.
In addition, an extension of the concept of information overload tested in this paper is to find the ideal number of categories in a product structure (when searching an entire web page) to determine the optimal level of choice at each level in the category structure and to the optimal number of levels in the web page structure.
We would like to thank Ravi Shanmugam and Yakov Bart and Roman at SisIQ for their extensive work and help in setting up website and study conditions through which the data for this paper could be collected. Thanks to Miguel Villas-Boas, Stefano DellaVigna, Noah Lim and Mario Capazzani for their valuable feedback.
- Jacoby, J., Kohn, C. A. and Speller, D. E. (1973) Time Spent Acquiring Information as a Function of Information Load and Organization. Proceedings of the American Psychological Association’s 81st Annual Convention, Vol. 8, pp. 813–814.Google Scholar
- Anderson, C. (2006) The Long Tail: Why the Future of Business Is Selling Less of More, Hachette Digital, Inc., New York, NY.Google Scholar
- Staelin, R. and Payne, J. W. (1976) ‘Studies of the information-seeking behaviour of consumers’, in Carroll, J. S. and Payne, J. W. (eds) Cognition and Social Behaviour, Laurence Erlbaum Associates, Inc. New York, NY, pp. 185–202.Google Scholar
- Iyengar, S. S., Huberman, G. and Jiang, W. (2004) ‘How much choice is too much? Contributions to 401(k) retirement plans’, in Mitchel, O. S., (ed) Pension Design and Structure: New Lessons from Behavioural Finance, Oxford University Press, Oxford, pp. 83–95.Google Scholar