In an era of endless choices, personalized recommendation services have become increasingly useful — and all the more present. From Netflix suggesting your next binge-worthy show to Spotify curating your perfect playlist, tech companies are vying to streamline and automate the consumer experience. But what makes some consumers more receptive to algorithmic recommendations than others?
Sonia Seung-Eun Kim, a doctoral candidate at the time of the study, and Professor Donald R. Lehmann, the George E. Warren Professor Emeritus of Business at Columbia Business School, set out to investigate this question. As the author of more than 200 articles and books, Lehmann has always been interested in the study of individual and group choice and decision making, making this research a natural fit for him.
Previous research in the field had indicated that people are generally wary of AI and algorithmic-driven recommendations, especially when it comes to making an emotional rather than a utilitarian choice about a product or service. “We were interested in identifying the type of consumers that would be more open to these personalized recommendation services,” says Kim.
Key Takeaways:
- People who frequently consume a wide variety of products or services are more receptive to personalized recommendation services.
- Perceptions that there are many different options to consume will make it more likely that consumers will be more open to services that provide recommendations, especially if it’s difficult for them to decide on an option.
- Companies can target consumers with high past variety or remind consumers of past variety to encourage trying recommendation services and choosing the recommended items.
The study’s findings: At the heart of Lehmann and Kim’s work is the concept of “past variety,” defined in this context as “the frequency with which someone tries or uses different products and services.” Using online platforms like Prolific and Connect, the researchers recruited over 1,000 participants, divided over four related studies, to test exactly why and when someone would entertain a recommendation from an app or similar service and how their personal consumption history played into their decision to do so.
Across their four studies, they repeatedly asked participants to gauge the degree to which they experienced past variety, framing this in domains like literature, music, fashion, and film. After analyzing the study data, Lehmann and Kim reached a clear, albeit surprising, conclusion: People who generally consumed the same things on a regular basis were less likely to consent to a new product trial or recommendation.
The researchers hypothesize that consuming a wide variety of products broadens a consumer’s sense of the options available to them in the market. Once aware of this range, they become more open to recommendation services that help them narrow down their options and find the one suited for them. The researchers established causation between past variety and openness to recommendations by randomly assigning study participants to either low- or high-variety groups. Participants were shown a week’s menu, with the high-variety group “consuming” a different entrée each day and the low-variety group getting a menu of only two randomly repeating meals. Participants were then asked if they would like to try a meal-recommendation app. Despite their limited culinary experience over the imagined week, the high-variety group was more likely to want to try the app. The same finding held when Lehmann and Kim performed the same experiment using clothing instead of food.
Moreover, even when consumers haven’t consumed a wide variety of products and services, it’s still possible to instill in them the sense that they have and see results comparable to those who have experienced high variety in the past. Lehmann and Kim demonstrated this effect by exposing two groups of participants to the same five songs. Although both groups were asked to rate (from 1 to 9) how much variety the song selection showed, only one group was asked to write a sentence about how each song differed from the others. This group later proved to be more open to trying a personalized music recommendation service.
Lehmann and Kim were also intrigued by the effect of decision difficulty and how this influenced participants’ responses. Participants who found it difficult to select a song from a list of songs, for example, were generally open to recommendation services regardless of the variety of music they consumed in the past. “If you perceive a task as being difficult, you’ll probably desire to narrow down your options,” says Lehmann, “even if you’ve experienced low variety.”
Why the research matters: Given the recent surge of interest in AI and the rate at which apps are proliferating, the applications of Lehmann and Kim’s work are vast. If consumers are more willing to try personalized recommendation services when they think they’ve experienced high variety, the trick is to make them more aware of the range of choices they’ve had in the past.
“One approach could be to let people know that, compared to the average person, they’ve had more variety,” says Lehmann. By helping consumers recognize that they are variety seekers at heart, companies can both target them and serve ads for services that simplify their search for the right product. “Additionally, companies may create ads that remind consumers of the amount of variety they have had in the recent past,” says Kim.
Adapted from “The Effect of Variety in Past Consumption on Openness to Personalized Recommendation Services” by Sonia Seung-Eun Kim and Donald R. Lehmann of Columbia Business School.
The effect of having variety in past consumption of a recommendation service is moderated, or mitigated, by the perceived difficulty of choosing a song. The effect of past variety is significant (p < :05) below the dashed line (3.71), while the effect is reduced for consumers who perceived song selection as inherently difficult.
