High-Value Verbiage Aids NLP in Predicting Consumer Behavior

Article Featured Image

A shotgun approach might work some of the time for the United States Marine Corps, but not, apparently, when it comes to ensuring its recruiting message hits the target.

In a campaign to identify an audience of potential recruits, with a focus on unemployed men from the Millennial generation, the Corps wanted better results than traditional targeting approaches were providing. So it decided to try something different, turning to a relatively unknown AI-driven tool called conversational analysis for a more refined, scientific approach to finding the right audience. The Corps used conversational data to build a new audience of some 2.6 million prime recruiting candidates with whom to engage, surpassing the campaign’s benchmark on-target rate by 139% Conversational analysis delivered that audience simply by identifying certain keywords and phrases young men were using in their online conversations, zeroing in on segments discussing college graduation, job search, unemployment, and other related topics.

Outcomes such as this hint at the massive — and still largely untapped — potential of conversational analysis as a marketing tool. With everyday human conversation shifting from face-to-face to online (we’ll let the social scientists and pundits weigh in on the implications of that phenomenon), there’s a virtually endless supply of digital dialogue for brands and marketers to try to make sense of. That’s where conversational analysis comes in. Using natural language processing (NLP), which employs artificial intelligence and machine learning to process and analyze large amounts of natural language data, conversational analysis parses the constant stream of online conversations to provide unprecedented predictive insight into consumer behavior.

Oracular Vernacular

There’s much to be learned about consumers’ wants, needs and emotions from the massive volume of digital dialogue spilled on social media. When people’s words come unsolicited and unprompted as they do in informal online conversations (versus via surveys, for example, which usually come with an inherent bias), they tend to carry greater weight, meaning and marketing value. Using advanced algorithms in tandem with NLP, conversational analysis does what other traditional targeting approaches — demographics, psychographics, purchase history, contextual or behavioral markers, etc. — cannot, contextually connecting what people say in their online conversations to how they’re likely to act as consumers.

Conversational analysis processes the constant torrent of online social chatter through the filter of an expansive taxonomy of keywords, phrases, and classifiers (audience.ai currently has a library of close to 35,000 of them, for example). That taxonomy is continually updated as new words — slang, jargon, brand and product names — enter the lexicon and others drop off. Conversational analysis plugs into that taxonomy to read between the lines of unprompted, unsolicited conversations to glean insight into how peoples’ use of specific words and phrases make them more or less inclined to buy a product. 

Unlike other targeting approaches, conversational analysis uncovers correlations in first-party data to read an audience’s emotions, needs, and intent. It removes much of the guesswork from targeting by predicting consumer behavior based not only on what they say but how strongly they say it. By tracing keywords back to specific conversations and user information, it creates profiles and defines unique audiences of motivated consumers who show the greatest likelihood of engaging and purchasing at the most opportune time (see the graphic below). 

By drawing from a living, breathing lexicon of words and phrases, conversational analysis recognizes nuance better that other types of targeting data, with the ability to extract both implicit and explicit meaning from those words and phrases when they’re used in an online conversation. In a case in which it was tasked to generate target audiences of people considering moving or renovating their residences, conversational analysis picked up on the implied meaning in comments like, "We are really outgrowing this place,” and "I hope mortgage rates stay this low” to create target segments of people who are seriously contemplating moving. Meanwhile, it also was building segments around more explicit language about mortgage brokers, school districts, etc., that suggests a move isn’t just being contemplated, it’s imminent. It also was able to define such segments as “soon-to-be empty-nesters” and “first-time home-buyers.”

This is exactly the type of relevant, segmented audiences that contractors, realtors, lenders, home improvement stores and other parties would love to reach. And in a marketing world that’s all about immediacy, conversational analysis, because it relies on fresh online conversations, would have enabled them to reach these audience segments as they were actively engaged in the process of, or in the contemplation of, moving or remodeling, rather than well after the fact. 

Conversation Starters

Conversational analysis is finding its way into the marketing mainstream as a growing number of brands turn to it for audience targeting. That includes Disney, which had great success using it for movie promotion, along with Hershey’s, Under Armour, Guardian Insurance and many others. Based on our observations of a range of campaigns involving conversational data, here are a few best practices to help maximize its impact: 

  1. Pick a subject about which there’s plenty of conversation. The greater the volume of fresh online conversations to analyze, the stronger, more targeted and more segmentable the conversational data, and the leads it yields, will tend to be.
  2. Test the waters. As new as conversational analysis may be to some organizations, and as scalable as the data is, it’s well suited to testing with a pilot program before scaling up based on results, budget and comfort level.
  3. Let conversational analysis stand alone. As tempting as it may be to overlay conversational data with demographic data from other sources, doing so lessens the clarity of the insight conversational data yields.
  4. Approach with an agile mindset and methodology. Conversational data sets are readily iterated and segmented in a multitude of ways, allowing the user to learn from their targeting experience on the fly and adjust accordingly.
  5. Use it for an online or offline campaign, or both. Conversational analysis can provide highly defined audiences for direct mail, call center, email, social media and other forms of marketing and promotion.

The buzz about conversational analysis is building in the marketing world. Is it time your company inserted itself into the conversation?

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues
Related Articles

Customer Support Predictions for 2020

Staying ahead of key trends can be key in helping companies create a personalized customer experience. Here are five key trends that will shape customer support in 2020.

U.S. Consumers Believe Customer Support Calls Put Them at Risk for Fraud

More than four in five Americans think brands could do more to prevent customer service fraud and one in four don't trust that brands are handling their personal information securely.

Speech Analytics Reveals Companies Aren’t F%!king Listening

As customer sentiment becomes increasingly angrier, businesses must consider the underlying issues before it's too late. So, what are the common triggers for the frustration, and how do we resolve them?