The Myth of Sentiment Analysis

One of the holy grails of marketing is figuring out a way to provide fully automated sentiment analysis, that magic algorithm that would give brands a way to instantly interpret the things users are saying about them online and make breakthrough business decisions based on the data.

But I’m not sure that’s happening any time soon.

To begin with, I’m suspicious of reading anything more into sentiment analysis than the broad stoke trends we already know how to find: does it matter if 74% of users don’t like your new product versus 86%? Not really. Either way, you need to fix the product. Sentiment analysis can show a trend—are a lot of people suddenly liking or disliking something? Is there any demographic pattern to who is liking or disliking most intensely?

But beyond that, I think we are fooling ourselves. To begin with, people posting online are arguably not a good sample of the population at large. It takes a certain personality type (call them “online extroverts”) to publicly post their opinion of anything. It’s less an age or culture thing as psychological: there are lots of people who are loathe to offer public opinions of anything in real life, let alone online.

“Public” is the key word here: among close friends, these same people will open up, the same way they do in what they perceive to be a “safe” (e.g. closed) social environment. And right now, the sentiment analysis scrapers can’t peer that far behind Mr. Zuckerberg’s magic curtain.

Sentiment analysis is also not sophisticated enough to pick up on cultural mores: expressing unfavorable opinions is looked on as bad manners in many cultures. (My wife is from Tennessee, and Southerners tend to favor the “damn with faint praise” approach. So if you hated “Black Swan” you’d say something like “that Natalie Portman has the most beautiful skin” or something equally benign. That’s not the sort of clue an algorithm is going to pick up on.)

Finally, there’s value in checking sentiment by hand. You can pick out things people are saying about your brand that you hadn’t even thought to have a program check. Plus you get a feel for the type of people who are talking about you.

But in the end, it’s all about the broad strokes: people rarely know why they really like or dislike something, they just know that they do.