UI Design Newsletter – September, 2006

Treating the customer right

Over the past few years, e-commerce and business-to-customer Web sites have seen a virtual explosion of interest (and investment!) in Web personalization applications and agents. These tools not only give the on-line merchant ways to understand customer preferences and purchase trends, but to apply them as personalized experiences for the individual consumer. E-commerce sites can then more effectively target purchase recommendations. When it works right, both consumers and merchants win: the shopper receives that personalized service they are looking for and the merchant is able to maximize their profits.

Is that what the buyer wanted to hear?

However, timing and presentation of product recommendations, and understanding of customer preferences are everything. The merchant must be able to match product recommendations with what the customer is interested in order to be successful. In addition, it's important to know whether the customer is thinking about or has already made a product decision, and is open to considering an accessory purchase, or whether the customer is at the very beginning of the purchase process and needs information on product alternatives. Lastly, the merchant needs to have that feel for when a "pressure" sale approach will help get the customer to buy their recommendation, or where a more positive presentation will help.

Negative consequences can be severe. In the best case, consumers will probably ignore any suggestions made by the recommendation agent in their final purchasing behavior. In the worst case, the consumer will leave the Web site empty handed and find some other site that better takes their needs into account.

What are they thinking?

A recent study by Ho and Tam defines an approach for good agent design. They used Consideration Set Theory (Roberts, 1989), to model the consumer's real-time decision process. Ho and Tam looked at:

• the timing of product recommendations,
• the influence of matching recommendations with consumer preferences,
• and the wording of recommendations,

to see how these influence consumer behavior.

Consideration Set Theory describes how consumers analyze, consider, and choose products; the theory has been studied since the 1980s (Andrews & Srinivasan, 1995; Gensch, 1987). There are four hierarchical sets or stages:

• Universal set: all possible product alternatives that could be purchased by the consumer
• Awareness set: a subset of all possible alternatives, limited by consumer interest and cognitive factors (attention, comprehension, cognitive capacity)
• Consideration set: small set of potential purchase alternatives, that the consumer carefully analyzes and compares
• Choice outcome: the best alternative that the consumer arrives at, after eliminating the other alternatives from the Consideration set.

In the study, the authors teamed with Hong Kong's largest data services company, who offers cell phone ring tone downloads as a service to their customers. The data services company provided historical ring tone purchase data for 7858 customers. This data was combined with current Billboard Music ratings to establish a prioritized pool of 72 ring tones to be used in the study. The ring tone list was used to determine participant music (singer) and music style (slow/fast beat) preferences and potential items for participant consideration and choice.

The study used a ring tone download Web application owned by the data services company. For each user session, a list of ring tone names were presented. For each ring tone, the participant could play the ring tone. Each ring tone listened to by the participant was counted as part of their "consideration set." Also, the participant was able to download a ring tone. The participant was only allowed to download one ring tone, and this was counted as their "choice outcome."

Before starting, participants filled out an on-line questionnaire, documenting music preferences and selecting and ranking their favorite 3 singers and music style (taken from the pool of 72 ring tones). This data was used to establish the recommendations that would be presented to each participant.

The authors looked at the importance of decision stage on considering and choosing an item recommendation, as well as the importance of matching customer preferences. Not surprisingly, agent recommendations that matched the participant's preferences were more likely to be part of their consideration set, and more likely to be the final choice. The timing of the recommendation was also important – the earlier the recommendation was made, the more it was considered and chosen.

Next the authors studied the "tone" of the presentation. Would it matter if the recommendations were made in a positive tone (e.g., "this item offer is just for you") vs. a negative one (e.g., "this is the last time you will get this offer")? And would the effect of being positive or negative vary depending on the purchase decision stage?

Recommendations worded in a negative tone had a greater impact on participant behavior – there was a greater chance for a recommendation to be considered and downloaded if it was worded negatively. And the effect of negative tone on recommendation downloads was most pronounced when the recommendation was presented during the choice outcome stage. Emotional tone had no effect when presented during the consideration stage.

A "personalized" opportunity

This study opens the door to providing actual research data in the areas of persuasion and influence in online shopping. Hard research in this area is still slow in coming, so the usability community welcomes any entries such as this. For now here is what we know and what we should do:

• Pay attention to when you use personalization. Think about the stage of the user decision process.
• Think about how to word your recommendations. In order to be effective you may need to "push" the user.

We look forward to the next set of studies in this relatively new field of research.

References

Andrews, R.L., and Srinivasan, T.C. (1995). Studying Consideration Effects in Empirical Choice Models Using Scanner Panel Data. Journal of Marketing Research, 32, 30-41.

Gensch, D. H. (1987). A Two-Stage Disaggregate Attribute Choice Model. Marketing Science, 7, 299-310.

Ho, S.Y. and Tam, Y.K. (2005). An Empirical Examination of Web Personalization at Different Stages of Decision Making. International Journal of Human-Computer Interaction, 19 (1), 95-112.

Roberts, J. (1989). A Grounded Model of Consideration Size and Composition. Advances in Consumer Research, 16, 749-757.

It seems like an obvious advantage to provide offers suited to the customer and present them at the "seducible moment" in the purchase cycle. It works in brick-and-mortar situations and it works on sites. But a negative tone is NOT common practice in most stores. Sure, there are some trendy shops that seem to select staff for bad attitudes. But most stores work to be positive. I suspect that there are undesirable side effects from negativity. You might make that sale, but have fewer people come back (which is NOT measured in the study cited). I would hesitate to use a negative tone but would love to hear more research.