Avoiding the dangers of online product recommenders
To help consumers navigate the unlimited choice of online retail, websites increasingly provide online productre commenders—tools that use consumer data to make suggestions about which products are best for them. Joseph Lajos uncovers the hidden dangers of thesetools for retailers and consumers.
“In the past 10 years, we have seen an explosion of product choices in almost every product category in the marketplace and especially online,” says Joseph Lajos. In response, websites increasingly offer online product recommenders to help consumers narrow down their choices. The tools make recommendations based on data about what other consumers with similar preferences have bought and liked previously (collaborative filtering) or by soliciting information directly from shoppers about how they plan to use products and which product attributes matter most to them (content filtering). How effective are these tools?What risks do they pose?
ONLINE PRODUCT RECOMMENDERS AND CONSUMER SATISFACTION
Previous research demonstrates that online product recommenders are highly effective in helping consumers identify leading product choices.Research also shows that consumers get more enjoyment out of the process of choosing when using these tools. “But there has been almost no work done on how online product recommenders influence satisfaction with the actual choices that people make,” says Lajos. “This is the focus of our project.” According to Lajos, this research was inspired by the work of social psychologist Timothy Wilson and, in particular, Wilson’s findings that over-analysis during decision-making can cause people to make less satisfying choices. “We identified this interesting finding from psychology and imported it to marketing, seeing a clean fit with online product recommenders.We hoped for the same results as Wilson but with an interpretation of the trade-offs between hedonic (emotional or fun) and utilitarian (functional) considerations.” Lajos explains that previous marketing research has established that consumers use both fun and functional considerations when buying products. Lajos and fellow researchers hypothesize that content filtering — asking consumers how they intend to use products and what kind of product features are important to them — causes people to make more functional choices than they normally would. And that, after a time, when people’s priorities have returned to normal, they may regret those biased choices.
RESEARCHERS DESIGN EXPERIMENT TO TEST ONLINE PRODUCT RECOMMENDERS
“We designed five experiments to test our hypotheses,” says Lajos, “and the simplest and most realistic of these is our trial involving nutrition bars — a product category we chose because of its equal split between functional and fun considerations.” The researchers created a fake website called Nutrition Zone, designing eight fake nutrition bar brands with two hedonic or emotional elements (packaging and photographs of people who supposedly endorse the brand) and two utilitarian or functional elements (nutrition facts and ingredient list). Four of the brands looked good but had low nutritional values, and the other four were low in aesthetic appeal but contained high nutritional values.
ONLINE PRODUCT RECOMMENDERS ENCOURAGE PARTICIPANTS TO CHOOSE FUNCTIONAL BRANDS
Participants were recruited off the street and divided into two groups. Believing they were evaluating a real nutrition website, the control group saw a list of the eight fictional brands on the Nutrition Zone website arranged in random order with no ratings and no information. They clicked through the brands to view descriptions (with photographs, nutritional information, etc.,) and make their choices. The second group answered content filtering questions and then saw the list of brands arranged according to the online product recommender’s rankings. This second group ultimately chose far more of the functional brands than the control group, confirming the researchers’ hypothesis that content filtering biases consumers to make more functional choices.
Bloggers transform the marketer's message to suit their own identity as well as their audience and their network
ONLINE PRODUCT RECOMMENDERS POTENTIALLY DECREASE CONSUMER SATISFACTION
Participants were given five of the trial bars they had chosen to take home. “But the secret was that everyone actually received the same bar, just packaged differently,” says Lajos. At a later date, participants filled out a follow-up survey. Even though the actual bars that everyone tried were identical, overall liking of the bars was much lower among the group who used the online product recommender (even in terms of taste!). “Somehow the process of answering the questions posed by the online product recommender biases participants to make more functional choices than they normally would. And, after time goes by, and they return to their normal mindsets, they end up regretting those decisions and wishing they had chosen less practical but more emotionally interesting options,” concludes Lajos.
IMPACT ON CHOICE SATISFACTION VARIES DEPENDING ON FUN VERSUS FUNCTIONAL TRADE-OFFS
“We conducted other experiments to examine the underlying processes more carefully. We wanted to make sure that we had identified what was really driving this effect,” says Lajos. Their findings were consistently confirmed in further studies. But when the researchers tested product categories that varied according to functional versus fun consideration ratios—one with a 50/50 split, another with a 70/30 split, and so on—they found that content filtering has no negative effect on consumer choice satisfaction in purely functional product categories. Similarly, products where emotional considerations are most important underwent the most severe decreases in consumer choice satisfaction.
For products that are more fun than functional, Lajos says the simple solution is to use collaborative filtering instead of content filtering, but not all retailers have the means to do this. “For Amazon. com, collaborative filtering is fine, because they have such a large bulk of consumer histories to work with, but for a smaller company or website, they may not have large enough consumer data banks to make collaborative filtering feasible.”
Based on an interview with Joseph Lajos and his talk, “Adverse effects of online product recommenders” (presentation of joint work with Amitava Chattopadhyay and Kishore Sengupta) at the CCIP Marketing Breakfast, 8 December 2011.
This research suggests it isimportant for business leaders to understand the trade-offs that consumers intheir target markets tend to make between functional and emotionalconsiderations. “If a product is primarily about emotional considerations andyou use an online product recommender with content filtering, consumers mightbuy your product in the short term, but they will be less likely to repurchaseor recommend them.” Lajos adds that marketers who absolutely must use contentfiltering can potentially overcome biases by designing questions that reflectpeople’s baseline priorities. “It is important to brainstorm ways to mitigatethese effects,” says Lajos.
The researchers conducted fiveexperiments using content filtering within a variety of product categories.They established boundary conditions via a test in which one group ofparticipants, after answering the content filtering questions, received anerror message that the recommender function was not working. This group stillchose more functional brands and was less satisfied later on with their choicesthan the control group that did not interact at all with the online productrecommender, suggesting that it is not the recommendations themselves that aredriving the effect in these studies but rather the process of answering questionsabout product attributes and use.