Why it comes to buying online, people are more likely to heed the opinion of Internet users in general rather than that of their own friends.

Researchers at the University of Columbia and at HP Labs set out to analyse the impact of general public opinion on the one hand, and that of friends on the other, on how a consumer makes up his/her mind to go ahead with a purchase. The researchers ran several surveys on Amazon's Mechanical Turk crowdsourcing platform between July and August this year. Those polled were asked to choose between two hotel reservations and then between two movie trailers. Each time they were also given some extra information: 1) the opinion of a large number of consumers, shown as a score of 1 to 5 stars, and 2) the number of friends recommending, or advising against, the choice in question. The report reveals first of all that an extra star on the general public score had a much greater impact on the user’s final decision than a friend’s recommendation.

Does public opinion count for more than that of a friend?

The study also showed that a negative opinion from an acquaintance or social network friend had a greater impact than a positive one with the same strength. So, one negative comment more or less will always count more than the presence or absence of a positive message. Thirdly, the researchers noticed that there were consistent variations according to the risk - financial risk in this case - associated with a decision. Booking a hotel costs money and making a mistake entails a lot of inconvenience, whereas a user only wastes a few minutes watching a poor film trailer.

Choice is more random if there’s no financial risk  

Responses were far more random (around 50-50) when users were watching the trailers than when they were booking hotels, and the impact of the opinions, either from the general consumer group or from the friends, was less clear-cut. The researchers believe that this study could help companies create more effective algorithms for the automated recommendation systems of online store sites and even advertising banners. However, they acknowledge that the study has its limits since it only shows the number of friends involved, but doesn’t identify them. In future they would therefore like to embed an identification parameter as they are convinced that the opinion of particular friends – family, people sharing the same interests, etc – might have a more decisive impact.