Your Customers Are Talking. Are You Listening?

Algorithm-based sentiment analysis is a game changer for customer experience initiatives.

By John Godwin

We see it with our clients and potential clients all the time: customer experience (CX) initiatives fall short or even fail to launch because their sponsors can’t build a case for change within the organization. What they need is a compelling storyline. Hunches and guesses, after all, are not enough to earn buy-in, secure funding and effect change. And when budgets are tight, skepticism flourishes. So where can you find compelling storylines to jumpstart your CX initiative? The answer is sentiment analysis.

Sentiment analysis is the collection and analysis of opinions about a product, service or brand. (Although it’s not exactly the same thing as emotion artificial intelligence — emotion AI — or data mining, people tend to use these phrases interchangeably and the subtle differences aren’t worth dwelling on here.) Traditional sentiment analysis can produce what feels like anecdotal evidence. It’s a time-consuming, monotonous task that’s prone to error. But machine learning’s ability to analyze large amounts of consumer opinion to reveal patterns radically transforms this process.

The more business interactions shift into the hands of the consumer — quite literally — and the more consumers use social media posts, product reviews and other expressions of opinion as their personal megaphone, the more important this kind of “super” sentiment analysis becomes. It is more accurate, faster and scalable.

More Accurate

Brandseye, a small firm based in Capetown that tracks real-time social media sentiment for the likes of Uber and Pizza Hut, used machine learning and sentiment analysis to do what polls and pundits couldn’t: predict Britain’s exit from the European Union and forecast Trump’s victory over Clinton. How did they do it? By using AI to scrape 37 million public social media conversations (read: consumer sentiment). To improve its accuracy, Brandseye crowdsourced human analysis of individual messages. (AI is limited when it comes to understanding context or detecting tone — sarcasm, hope, etc. — but there are, as this example illustrates, ways to compensate for this.)

Faster

In 2014, machine learning quickly enabled Expedia Canada to see that more than half of the people commenting on its “Escape Winter: Fear” commercial, which had performed well in focus groups, hated the violin soundtrack. Meant to be slightly annoying, it became insufferable for those who saw it play in “bonus” spots over and over during the World Junior Hockey Championships.

When the Twittersphere demanded revenge, Expedia complied.

The company created three new videos in response, including one where an angry citizen who had Tweeted that someone should smash the shrill violin got to do just that. “We’re listening,” the voiceover declared. “And hopefully that’s music to your ears.” Expedia’s ability to pivot quickly — and to do so with a sense of humor — was a win for the brand.

Scalable

Did I mention that Brandseye scraped 37 million public social media conversations? That’s six zeroes, folks.

Social media conversations — about both you and your competition — are a gold mine of consumer sentiment.

But what else can companies study with algorithm-based sentiment analysis?

• Product reviews
• Survey responses
• Web pages and forums
• Call center transcripts

To get started, you have to choose the best model and toolkit for your purposes.

Here are a few of the questions you’ll need to answer, probably with the help of a customer experience expert:

• Which is the best algorithm to capture the info you need?
• How do you identify the correct phrases to analyze?
• How do you convert the insights you gather into better products, services and experiences for your customer?

Sentiment analysis can help you identify:

• Areas of strength (so you know what’s working) and opportunities (so you know what isn’t). The latter are your opportunities, some of which can be surprisingly low-hanging fruit.
• Training opportunities for call center and help desk employees or other agents of the company. When working with a Fortune 100 property and casualty insurer, sentiment analysis led us to suggest customer service reps address callers by name and ask the caller whether they are in a safe location before starting a conversation. Sometimes a simple tweak on the CX frontline is all you need.
• Themes for new sales and marketing campaigns. Because sometimes you have to smash the violin, no matter how well it performed in focus groups.

Keep in mind that though I’ve talked about sentiment analysis as a great way to jumpstart a CX campaign, this should be an ongoing effort. Continuous sentiment analysis fuels an iterative test-and-learn strategy, the agile approach we should all be taking.

No matter what your goals, informed business decisions are better decisions. And the availability of algorithm-based sentiment analysis means there are no more excuses for not knowing how your customers feel about you. Start listening: it’s music to their ears.