Sentiment Analysis Can Help Find the Idea in That Pile of Data

Published on .

Katy Keim
Web analytics have unleashed a monster. I believe strongly in outcome metrics, crystal-clear measurement and analysis for improvement. It is critical for the operational aspects of marketing.

But as I sat at the Argyle CMO Forum in New York City back in January, I felt a tinge of regret as marketers indicated that measurement was at the top of the priority list but, at the same time, admitted they were under "an avalanche of data."

The leader of United's Mileage Plus program rattled off all the key metrics for his business -- program participants, miles redeemed, and the effective currency rate of mileage. Metrics? He had a river of them. Loyalty? Not sure he had much.

Mr. United Loyalty didn't mention the word "customer" once.

In the margins of my handy conference agenda I wrote: "STILL ABOUT THE BIG IDEA." And the big idea for marketing is to connect customers with a company.

On the other hand, the Q&A couldn't go on long enough with Kerry Holland Tillman from Heineken, as she explained the strategy behind the "Most Interesting Man in the World" campaign for the company's Dos Equis brand. As you know, the tagline in the commercial is : "I don't always drink beer, but when I do, I drink Dos Equis." Management was initially up in arms about a beer spokesperson admitting he doesn't always drink beer. Yet this approach was one that created an authentic connection with the audience. It was smart. It was bold. But how in the world would you have justified it with numbers?

Here's one way to commit to measurement while honoring your spirit, your customers, and the big idea: Measure what your customers say about you when they're talking to each other.

Some technology vendors call this social-media monitoring (guilty!). That's a terrible term. You're not monitoring media, you're listening to your customers. And what can you find out?

Are your customers more positive about you than they were last month? Are they asking more of the kinds of questions that mean they're considering purchases? Or switching to a different brand? What other topics do they talk about when they talk about you? Are there abbreviations or slang associated with your brand? And which of those topics do they really care about?

I know the objections to sentiment analysis and natural-language processing. This kind of measurement isn't precise enough. Sometimes people are sarcastic. The words people use change all the time. There are ambiguities too: "Read the book" is a positive comment for a literary review but a negative comment for a movie review.

Technology will bring you to the data but you have to drink it, absorb it and do something valuable with it. Give it context. Sentiment data is repeatedly tested and benchmarked against human agreement, but make no mistake, there is no substitute for your own insights.

But the big idea isn't about the difference between 75% and 81% on a sentiment graph. The big idea is about spotting a trend faster than everyone else, and then finding meaning in it.

And when you can see a trend line going up, then look in real time and see the customer conversations that are driving that trend that 's more powerful than raw numbers. It might give you a big idea -- or help you sell one you already have.

Show me an idea that has been projected, measured, processed, flowcharted, spread sheeted, analyzed, and stripped of its customer context, and I'll show you a small idea.

You don't get to skirt sound measurement, but you can ask for a new dimension. A human dimension. Spirit.

Katy Keim is the CMO of Lithium Technologies.
In this article:
Most Popular