3 ways data science is driving text analytics innovation

Dennis Ehrich | Feb 12, 2019 Dennis Ehrich 02/12/19

If you work with data, you’ve probably noticed the phrase “data science” being thrown around a lot more frequently as of late. While terms like machine learning and artificial intelligence (AI) may feel like they’re bordering on buzzword territory, the hype is warranted. A recent study by Deloitte revealed that 88% of companies surveyed plan to increase spending on cognitive technologies in the coming year.

What’s driving this widespread adoption is the practicality and breadth of data science applications, specifically when it comes to automating routine tasks—which frees up teams for more strategic work with value-add potential. To illustrate this, we’ll break down how SMG is using data science techniques to drive text analytics innovations that help clients get more from their open-ended comments.

  1. Machine learning algorithms improve + continuously fine-tune sentiment accuracy

    For customer experience (CX) professionals, the biggest challenge with qualitative feedback is that it comes in such massive volumes, it’s impossible to comb through each comment for insight. But at the same time, each comment can help answer questions you might not have thought to ask. To solve this, text analytics engines work to automatically pull out common themes and qualify whether the comment bears positive or negative sentiment.

    While those out-of-box engines help save time, they’re rigid enough that they often require manual manipulation to work accurately. But with machine learning algorithms, SMG’s data science team can train the engine using hundreds of millions of historical comments from industry-specific text benchmarks. As the sentiment accuracy is continuously fine-tuned by new input, teams can spend less time monitoring the conversation and more time searching the relevant data for insight.


  2. Deep learning models help companies cast a wider net for comprehensive analysis

    Another major hurdle with open-ended feedback is the broad range in phrasing used by respondents. Consider a retailer that has thousands of SKUs categorized into different product types and spread across multiple departments. Beyond having to painstakingly add these terms to build out a library of entities unique to the brand, the effectiveness of that library is entirely dependent on whether customers use those specific terms.

    Fortunately, deep learning—a more sophisticated type of machine learning—is optimized for complex problems because it uses frameworks of multiple, layered algorithms working in concert. And by adding structure through linguistic hierarchies, all mentions are categorized appropriately (i.e., variations of product mentions are coded to the appropriate categories/departments for drill-down and roll-up analyses). Most importantly, by helping brands capture more product mentions with less manual effort, CX teams can turn their attention to evolving product strategies by:

    • Gauging the success of limited time offers + new services in real time
    • Tracking frequency of mentions to forecast demand + monitor trends
    • Prioritizing action items by calculating impact to Overall Satisfaction + sales


  3. AI capabilities help brands identify + alert on low-occurrence, high-impact events
    While increasing accuracy and comprehensiveness helps dial up the efficiency of text analytics solutions, there’s still one major aspect of CX insights we haven’t touched on: timeliness. This becomes even more important when considering events that require your immediate attention—things like foodborne illness, major operational issues, or employee misconduct. While these types of events may only appear once in tens of thousands of comments, a single instance is enough to make headlines.

    That’s why SMG’s data science team is training the text analytics engine to not only identify these rare instances in real time, but also use AI capabilities to trigger the appropriate actions autonomously. By using industry benchmarks to determine the scope of the issue, the model is able to determine whether to send an alert to the manager containing relevant information around coaching opportunities or diagnose and relay potentially system-wide issues to the company’s central customer care center.

Cognitive technologies offer enormous potential for CX professionals

There’s no question leading organizations are investing heavily in data science—and for good reason. By working tirelessly to put actionable data in the hands of those who can take action, data science techniques are dramatically increasing speed-to-insight capabilities and enabling organizations to be more agile in adapting to opportunities as they arise.

To learn how SMG is using cognitive technologies to help clients inspire smart changes across their enterprise, download our white paper: 3 ways data science is reshaping how brands approach CX data.

Dennis Ehrich
Chief Product + Technology Officer

Customer Experience Update