Ratings we assign are influenced by other ratings that we read. New research conducted by Sinan Aral and colleagues at MIT’s Sloan School found that:
“When it comes to online ratings, our herd instincts combine with our susceptibility to positive ‘social influence.’ When we see that other people have appreciated a certain book, enjoyed a hotel or restaurant or liked a particular doctor — and rewarded them with a high online rating — this can cause us to feel the same positive feelings about the book, hotel, restaurant or doctor and to likewise provide a similarly high online rating.
This important finding was discovered too late to be included in the Field Guide’s entry on “Influencers.” That discussion pointed out the importance of the herd model and urged that it be considered along with the widely adopted Influencer-Follower (two-step) model. The influencer two-step is very popular because: a) it conforms to the conventional mental models we evoke to explain how advertising works (authority, message, persuasion), b) because measures of influence, such as Klout scores, are computed in line with the two-step model — using social media counts such as posts/updates, number of friends/followers/contacts, and sharing, and c) herd influence has been under-recognized.
Despite books and articles on herd instincts in marketing, knowledge about herd instincts and its applicability to the work we do is not yet generally known by practitioners. Adding to herd instincts’ invisibility: Herd instincts measures are not reported by measurement services, so most of us are unaware of the herd notion. Herd measures are not easily derived from social media metrics. Methods for researching herd instincts scientifically in marketing and advertising are not in the market research tookit. This study changes that at last … and to our benefit.
Key Implication: Brands should oversee their ratings sections to minimize fraudulent positive ratings. Those “false positives” can create unrealistic expectations. Instead, encourage people to record authentic ratings that minimize bandwagon effects and foster realistic expectations about the brand. Ratings may then generate better guidance to other readers and to the brand itself.
The study had three experimental conditions, one where the rating was increased positively, one where the rating was decreased negatively, and one with no change. Read on for the five key findings …
- The positive manipulation increased the likelihood of positive ratings by 32% and created accumulating positive herding that increased final ratings by 25% on average.
- Positively treated comments were also significantly more likely than those in the control group to accumulate exceptionally high scores. Up-treated comments were 30% more likely to reach or exceed a score of 10. (The mean rating on the site is 1.9.)
- Positive social influence and negative social influence created asymmetric herding effects. Negatively treated comments received down votes with a significantly higher probability than control comments. But this effect was offset by a larger correction effect, whereby these comments were up-voted with a significantly higher probability than were the control comments. This correction neutralized social influence in the ratings of negatively manipulated comments.
- Herding effects and ratings bubbles varied by topic, implying that some product or service categories are more susceptible to social influence bias than others. While comments on business, culture and society, and politics were highly susceptible to popularity bubbles, those in general news, IT, economics and fun were not.
- Friendship moderated the impact of social influence on rating behavior. Friends tended to herd on current positive ratings and to correct comments that had negatively manipulated ratings, while enemies’ ratings were unaffected by our treatment. (However, this could be because of the small sample of potential first ratings by enemies. Though there are a substantial number of enemies in the community, they were less active, yielding a smaller sample of enemies’ ratings.)