CHI 97 Electronic Publications: Late-Breaking/Short Talks

Effective Product Selection in Electronic Catalogs

Patrick Steiger and Markus Stolze
IBM Research Division, Zurich Research Laboratory
Säumerstrasse 4, CH-8803 Rüschlikon, Switzerland
+41 1 724 8942,
+41 1 724 8263,


Product catalogs are crucial for electronic commerce on the Internet, but it is still a challenging task for casual users to perform effective product selection. Recently, a promising technique for product selection has been proposed: Incremental restriction on interactive tables. It allows customers to build complex queries with a few mouse clicks, but still to browse the available products at any stage. This paper describes effective and ineffective strategies of users working with this technique. These strategies were identified in a study with casual users.


Electronic Catalogs, Product Selection, User Studies

© 1997 Copyright on this material is held by the authors.


In the realm of electronic commerce on the Internet casual users are increasingly confronted with electronic product catalogs. There are three important techniques for product selection: browsing, database retrieval, and incremental restriction on interactive tables.

Browsing is usually the easiest searching technique for casual users. If they are used to browsing the web, they know how to browse a product catalog. It is easy to explore what is available. Unfortunately, if they know in advance what they are looking for (e.g. a list of criteria the product must fulfill) or if the list of products is long, it is cumbersome to browse forward and backward to compare several product features.

Database retrieval perfectly serves users who know exactly what they are looking for. Users enter a specific query reflecting their needs and they are, in return, presented with a list of appropriate products. However, if users do not know what they are looking for, it may be difficult to proceed. Data exploration has to be done by formulating several clever database queries. This can be a prohibitively complicated task for casual users.

Incremental Restriction on Interactive Tables (IRIT) is a new technique introduced by Spenke et al. [3]. It extends Ahlberg and Shneiderman's dynamic queries [1], and Rao and Card's focus+context technique [2]. IRIT combines the advantages of browsing and database retrieval: Starting with a potentially very large table, the user can incrementally restrict the view on the table to a relevant subset of products by selecting the criteria those products have to satisfy. After each point-and-click operation the user can browse the remaining subset to get a feeling of which products are available that match the criteria given so far. Spenke et al. implemented the IRIT technique in a tool called FOCUS [3].

We studied casual users working with FOCUS to better understand the impact of the promising IRIT technique on product selection effectiveness.


FOCUS [3] is a Windows 95 application developed at GMD, the German National Research Center for Information Technology. The software and documentation is freely available at

The interface is spreadsheet-like; columns represent products, rows represent features (product attributes). All available products are visible in the given window, although if there are many products, the columns may have a width of just a few pixels. If users want to inspect a single product, they click on the header of the corresponding column and the column is expanded to a readable width. When users select a feature (row) within this column, all other products that have the same value for that feature are also highlighted. By double-clicking on that cell, the view is restricted to those products that match this criterion. There are several other `query building' manipulations such as sorting the products by a feature or specifying a subrange of a feature for restriction.



Our test hypothesis was that users might apply a sequential top-down selection strategy, which is potentially suboptimal. This strategy implies that users restrict the number of visible products by deciding on their need for certain features in sequential order as they are listed in the table.

This strategy would lead to the selection of suboptimal products in situations that are similar but more complex than the ones shown below. In Figure 1 the application of a sequential selection strategy implies that, first, feature A is considered to be relevant and is therefore selected. This operation excludes product 2 and leaves only product 1 which, however, does not offer feature B. This is a sub-optimal choice if feature B is more important to the user than feature A.

Product 1

Product 2
Feature A

Feature B

Figure 1: Simplified FOCUS product table showing two products, each of which supports (X) one feature.

Starting from the situation in Figure 2, a sequential top-down strategy means that the user first considers A useful and selects it, thereby excluding the cheap product 2. Next, feature B is selected because it is considered even more important than A. Only the very expensive product 3 remains. The user does not become aware of the choice between obtaining both features A and B for $500 or only B for $100.

Product 1

Product 2
Product 3
Feature A

Feature B

$ 200
$ 100
$ 500
Figure 2: Simplified FOCUS table in which each of the products supports one or both features. The high price of product 3 is because it has additional features not important to the user and not shown here.

Test Procedure

Three men and five women, all employees of the IBM Zurich research lab, participated in our study. All of them were familiar with basic GUI applications, but none of them had had previous exposure to the FOCUS tool.

In the test session each subject received a brief demonstration of FOCUS. Then they had to perform a short warm-up task with a product table that was unrelated to the later task. After this, the subjects used FOCUS to select an appropriate travel insurance for a specified trip. The product table contained 32 products with 29 features in 3 main groups. Of these features, 18 had simple yes/no values. The remaining 11 features were primarily numerical descriptions of insurance coverage limitations.

During the task, the sequence of selection steps was manually recorded. Help was given to anyone who had questions concerning the operation of the tool and the meaning of product features (i.e. insurance services).

Once a product had been selected, the subject was asked to rate the importance of each of the product features on a separate sheet of paper. This rating was then used to assess whether participants selected the product that optimally matched their preferences.


Five out of eight people selected a product that matched their preferences and their budget. Two of the remaining three subjects found a product that matched all their preferences, but at a very high price (according to the example in Figure 2). One person ended up selecting a product that was not optimal with respect to the stated preferences.

None of the subjects relied exclusively on a sequential selection strategy, but most used sequential selection as a micro-strategy during part of their problem-solving process. Upon reaching a desired feature in the list, they selected it without considering that there might be other product features further down the list that they would have rated more important. If this problem was not detected before the next selection was made, users were led into dead-end situations from which many participants needed help to back out of. The three suboptimal solutions can be directly attributed to sequential selection. Effective users avoided getting into problematic situations by using a combination of the following strategies:

We further observed that many users wanted to be able to express their indifference toward a specific feature.

In conclusion, our study showed that some users do have difficulty working with the IRIT technique, but that there are strategies to avoid these problems. We are planning to explore different means to enhance and facilitate the use of effective product selection strategies by casual users. In particular, we want to investigate strategy instruction, demonstration and training of additional FOCUS features, as well as supply special system support.


We appreciate the excellent work of M. Spenke, C. Beilken and T. Berlage on the FOCUS tool.


1. Ahlberg C, Shneiderman B: Visual Information Seeking: Tight Coupling of Dynamic Query Filters with Starfield Displays. In Proceedings of CHI '94, Boston, 1994, ACM Press, 313-317.

2. Rao R, Card S: The Table Lens: Merging Graphical and Symbolic Representations in an Interactive Focus+ Context Visualization for Tabular Information. In Proceedings of CHI '94, Boston, 1994, ACM Press, 318-322.

3. Spenke M, Beilken C, Berlage T: FOCUS: The Interactive Table for Product Comparison and Selection. In Proceedings of UIST '96, Seattle, 1996, ACM Press, 41-50.

CHI 97 Electronic Publications: Late-Breaking/Short Talks