Are you more likely to see a movie or buy a bottle of wine if someone recommends it to you? Would you feel as strongly about your choice if the reviewer simply told you that he or she liked the product, but didn’t use the word “recommend”?
New research by Wharton marketing professor Jonah Berger reveals some surprising data about how consumers react to specific language used in reviews and recommendations, and whether or not that language could lead some to make bad decisions. His paper, coauthored by Grant Packard from the Lazaridis School of Business and Economics at the Wilfrid Laurier University in Ontario, Canada, is titled “How Language Shapes Word of Mouth’s Impact.” Berger recently spoke with Knowledge at Wharton about the paper, and offered some advice about heeding word of mouth.
An edited transcript of the conversation follows.
Knowledge at Wharton: Your previous work has looked at how content goes viral. Your latest book is Invisible Influence: The Hidden Forces That Shape Behavior, which looks at how hidden influences affect our decision-making processes. Could you summarize what you and your co-author, Grant Packard, looked at in this paper?
Jonah Berger: We’ve all seen the power of word of mouth, whether we’re making a simple decision like what breakfast cereal to buy, or a more important one like which house to choose. We use online reviews and word of mouth all the time to help us make those decisions. But is that word of mouth always helpful? That’s really what this research looks at.
Imagine you’re at a party or a conference, for example, and you’re talking about movies with two people you haven’t met before. One person says they like movie A, and another person says they recommend movie B. Which of those movies are you more likely to see as a result, and are you going to be happy with your choice?
“Knowing that a given movie or a given brand has a certain number of likes isn’t as impactful as knowing it has a certain amount of recommendations.”
What we find is that people are more likely to follow recommendations. You’re more likely to see movie B because you think the other person liked it more and it’s a better movie. But you might end up not liking that movie so much. You might end up making a worse choice because of the type of people that tend to use the word “I recommend.” It’s a language device that suggests not only you like something, but also you’re making an inference about what someone else likes. Whether we’re looking at books or wine or hotels, novices are more likely to say that they recommend something than experts. Experts aren’t as willing to use that “recommend” language. They’re more likely to say “I like” something. But they’re less willing to make a guess about what you’re going to like. As a result, if people end up listening to recommendations, as they often do, they might sometimes end up making worse choices.
Knowledge at Wharton: What are the key takeaways from your study?
Berger: We look at two things. One is how people endorse things. Sometimes people say “I like” something. Sometimes they recommend something. Those might seem like really subtle differences in language, but they have a big impact on two things. First of all, whether we’re persuaded by that language. Do we take that person’s endorsement and end up going to that movie or that restaurant? Also, whether we end up making a good choice as a result. Lots of research shows that word of mouth is really helpful. And indeed, often it is. But in some cases where we don’t know if someone’s an expert or not — like two people we meet at a conference that we’ve never met before — should we take their advice or not? When we can’t tell if someone’s an expert, or they’re not our best friend who knows a lot about movies, sometimes we use what they say and the way they say it as a cue to whether they’re an expert or not.
We assume that if someone says “I recommend” something, they actually know a lot about movies, for example. Or if they just say, “I like it,” we assume they don’t know as much. What’s dangerous there, though, is that the opposite is true. Novices, people who don’t know a lot about movies, are more willing to say they recommend something. The same thing happens with restaurants or other domains because they don’t think about the fact that others may have different preferences than them.
If you’re an expert, you’re not really willing to say “I recommend” something. If we don’t know each other well, I don’t know your taste. I don’t know if you like the same movies that I like. So, I’m not as willing to recommend it for you. But if I’m a novice, I’m very willing to use that strong recommendation — to say, “I recommendation this movie.”
Knowledge at Wharton: So, my recommendation could mislead somebody?
Berger: Exactly. In general, if we follow experts, that’s a good thing, right? When we look online, we follow the wine experts when we’re picking wine. We look for reviewer badges to figure out who knows a lot. But there are many cases in our lives where we don’t know whether someone knows a lot or not, so we use their language as a cue to whether they have expertise. But that cue may sometimes lead us astray.
Knowledge at Wharton: Did the findings surprise you?
Berger: Definitely. We [already] thought that recommendations would be more impactful than likes — that someone saying “I recommend this movie” seems stronger. It suggests, one, that they know a lot about the domain. But also, they just plain liked it more. So, that didn’t surprise us so much.
What did surprise us a little bit is the type of people that use these different types of language. I’ve done a lot of work on word of mouth. In general, I think word of mouth is a good thing. But in this case, sometimes word of mouth can actually lead us astray, because different people tend to use different types of language.
“What you don’t want to do is encourage people to follow the wrong information.”
Knowledge at Wharton: What do you think the key takeaways would be for marketers?
Berger: For marketers, I think it’s interesting. The first thing that probably comes to mind when people think of “likes” versus “recommendations” is Facebook — where we say, “Hey, I like this.” If they instead changed it to “I recommend this,” it would probably have more impact. Knowing that a given movie or a given brand has a certain number of likes isn’t as impactful as knowing it has a certain amount of recommendations. By subtly changing the language you use, you can impact whether people follow that language or not.
That said, you need to be careful. What you don’t want to do is encourage people to follow the wrong information. We did a study where we show that novices end up picking worse wine, for example. They end up being more likely to recommend that wine. And then following their word of mouth leads people to be more likely to choose bad wine over good wine, if they don’t know the difference. To be careful, marketers need to think about, “Are we allowing people to use other cues as well? Are we just putting the language out there, or are we giving information about how many books that person has reviewed? How many other bottles of wine they’ve talked about before? Or potentially other cues that will allow people to get a better sense of whether this is good for this person, or might it be particularly good for me?”
Knowledge at Wharton: What about the consumer side? What are the implications for people who are trying to get information about what they like or what they may want to buy? How can they judge a source?
Berger: Next time someone tells you they recommend something, be very, very careful. First thing, think about whether you know them or not. If it’s someone you know, and you know their preferences, and they know your preferences, the fact that they’re recommending something’s probably a good signal. But if you don’t know them, if you’ve just met them, if you don’t know whether they know a lot about a domain, a red flag should go up. If someone is saying they recommend something, be a little bit more careful. Figure out whether they actually know something about that domain. Do they know something about me? Make sure they know enough about your preferences before you go ahead and take that recommendation.
Knowledge at Wharton: I can see this playing out all the time on Facebook, for example. When it’s your close circle of friends, it makes perfect sense to accept an explicit recommendation. But I guess the caution comes in when it’s someone you may not know.
“Next time someone tells you they recommend something, be very, very careful.”
Berger: Yes. And I think a lot of times when we’re talking face to face, it’s pretty clear. Online, the boundaries blur a little bit. We talk to our good friends, but we also talk to a lot of people we don’t know as well. We often assume those people know a lot about what they’re talking about. They may not, necessarily. That, I think, is what’s important here. Lots of work has been done on the power of online reviews. They’re certainly a valuable tool to often help us make better and faster and easier decisions. But on the margin, there are some cases where it can lead us astray. And that’s what I think this research points out.
Knowledge at Wharton: What research topics do you want to look at next?
Berger: I think one thing this paper starts to do — and researchers are starting to look in this area more generally — is the language people use when they describe things. You would think in marketing that we’ve thought a lot about language — the language that works in advertisements, or why certain language would be more effective in word of mouth. We really haven’t that much. There’s a lot of work now bridging marketing as well as computer science — called natural language processing — that’s starting to look at not just whether people recommend something but what they say in that recommendation. Even in saying, “I recommend” something, I can say, “I strongly recommend this” or “I recommend this for people who like this type of other thing.” The words I’m using there can be quite important.
While a lot of research just looks at the number of stars [in reviews] — for example, saying a five-star review on Amazon leads to about 20 more books being sold — what’s also important to ask is, what’s the language being used in that review? How do certain types of language affect [consumers]? We’re going to do a bunch of research on this. We’re looking at the language of customer service calls. If you call up a retailer and say, “I’m really unhappy,” what words do they use that might make you happier, or not as happy?
We’re doing some work on the language used in movies, for example. Might the certain words people use over time in movies lead those movies to be more or less successful, and can we predict that based on the language of those movies? And song lyrics: Can we predict how successful a song is going to be based on the lyrics contained in that song? There’s a lot of interesting data out there that’s textually based, and we’re trying to understand it better.
