Zoom in on Online Choice

KIM, Jun | BRONNENBERG, Bart J. | MELA, Carl F.

Price is usually the main objective of search for homogenous products such as books. However, considerably less is known about the more complex process of searching for differentiated products such as consumer durables goods. Bart J. Bronnenberg, Jun B. Kim and Carl F. Mela set out to close this gap by studying online consumers’ search behavior for digital cameras. They study how consumers search for relatively expensive and differentiated durable goods, characterizing the extent of the search, the degree to which search is informative of choice, how the search changes over time, and which strategies of search consumers use to find decision relevant information. They achieved this using a very large scale data set that captured a detailed level of consumer search and attribute information.

They found that online searches for digital cameras are extensive across products, sellers, and time. Prior to buying a digital camera, consumers conducted an average of 14 online searches, navigating over three brands, six models and four domains during six sessions spread over two weeks. Even with extensive search activity, the search was confined to a surprisingly small product area and the camera chosen almost always lies in the “subspace” spanned by cameras searched but not chosen. This is true even during early searching, and more so during late searches. Furthermore, it is common for consumers to revisit alternatives that have previously been searched. Nearly a third of all searches are revisits and such behavior is strongest for alternatives ultimately chosen. About 70% of consumers search the camera they purchase online at only one online retailer. It is thus not common for consumers to first choose a specific camera model and then “price shop” this camera across retailers. By contrast, it is very common for a consumer to search different cameras at a given retailer.

Consumers “zoom in” on attribute levels (categories of product features) from both above and below.  At the same time, the mean attribute level searched remains close to constant throughout the search. Other dynamics are evident during the search. The chosen alternative is typically “discovered,” i.e., first searched, late in the search. Next, consumers tend to use more generic keywords during the earlier search phase and shift to specific keywords later. Finally, the calendar time lapsed between pages searched accelerates as choice approaches: in other words, searching the first camera takes more time than searching the last camera.

The fact that preferences revealed during the search are highly similar to those revealed by choice implies that the search is useful in identifying preference.  This information can be used by marketers to better anticipate purchases and make recommendations that result in better outcomes for consumers and firms. Firms can presumably use early search histories to target messages and better recommend alternatives as search proceeds.