Why Robots Shouldn't Replace People For Social-Media Monitoring

Algorithms Can't Tell The Intent Behind a Twitter Post. So Stop Asking Them To Try.

By Published on .

Darrell Jursa
Darrell Jursa
By now, everyone agrees that social-media channels like Facebook and Twitter provide marketers a useful means to talk to their customers. And I suspect most would agree that being able to listen in on conversations about our brands and products is even more important.

After all: You wouldn't show up to a party and start blathering at other guests until you had a pretty good feel for the tenor of the conversation taking place before you got there, would you?

So it's a little strange that so many marketers and agency executives who agree listening to social-media conversations is crucial have outsourced the job to the technological equivalent of those contextual banner-ad services that put airline ads next to plane-crash stories.

Don't believe me? Consider these results from a well-respected and widely used automated online-sentiment measuring service, which we asked to tell us what people were saying about Starbucks.

Here's one it thought was "negative" about the ubiquitous coffee chain: "im tired of this cheap coffee crap. need starbucks!"

And one it thought was positive: "Pumpkin chai. Mmm. Very good. Not from Starbucks. From Argo Tea."

Needless to say, marketers who are relying on unchecked data based on those conclusions are flying blind – or worse. It's not because the technology they're using is poor. (It's actually quite good at doing what they promise, which is to deliver a 65% accuracy rate.) It's because algorithms and keyword-searches are not a substitute for social skills.

So what's a digital marketer to do? They should start by taking a page from that most analog of marketing tools, the focus group, by using those new listening tools to put a representative sample of reactions and sentiment in front of an actual human being (via a dashboard that finds and presents the raw chatter). This is really the only way to understand the context of what's being said and why.

Yes, it's more labor-intensive than an automated approach. And there are potential hurdles –like determining a consistent definition for what constitutes positive and negative—that will need to be cleared getting started.

But an actual person isn't going to misinterpret the Tweets above the way automated services did, and they're far more likely to grasp the context surrounding those discussions, meaning that marketers are far less likely to embarrass themselves if and when they finally decide to join in the conversation.

Darrell Jursa is VP-Emerging Media at Dig Communications, and a regular contributor to the agency's blog, The Think Room.
In this article:

Comments (17)