7 Ways to Make Sure You’re Getting the Most of Your Unstructured Data with Text Analytics Today and Tomorrow
Yesterday my 10-year-old daughter sent 43 text messages, posted 8 Musical.ly videos, and put up two new Instagram posts. And given the fact she spent 8 hours at school and 4 hours at gymnastics practice, that would be considered a “lite” day with technology. Contrast that to the amount of digital content I shared as a 10-year-old. Yes – zilch, noda, nothing.
While there is a tremendous amount of digital content out there today, my daughter represents the up and coming generation of prolific social sharers who have known nothing but that sharing behavior. Recent estimates expect the volume of digital content to grow by a factor of 10 to 44 zetabytes by 2020. I can rest assured that my daughters will be doing what they can to make that estimate a reality.
Between the explosion of digital content and the rise of cloud computing, the amount of unstructured data generated by customers that we will be trying to understand in coming years will become even more daunting than it is today.
So the short story is the embracing and excelling at text analytics while critical today, will be an imperative tomorrow.
So what are a couple of ways we can take text analytic efforts to the next level? We talked about this topic a lot at the recent CXFusion conference. Here are 7 ways we talked about making sure your text analytics can keep up today and set the foundation for success in the even more prolific generation of tomorrow:
- Ensure you get rich and detailed information from customers by asking the right questions and probing for detail. Bottom line, if you don’t have rich meaningful data, you aren’t going to get much out of the data. When we converse in our daily lives, we ask questions and probe to learn more from those around us. Consider the using a technology equivalent of that behavior such as SmartProbeTM which will help to probe for more detail from customers or even probe on specific concepts that customers bring up in their comments.
Also, be sure to structure your open-ended questions appropriately to maximize sentiment accuracy. For example, one of the most common offenders is asking a double-barreled question. . . “What did you like or dislike about your experience?” When the respondent answers “the salesperson” sentiment becomes a mystery we can never solve.
- Make sure you work with a provider that excels in data preparation and cleaning. Yes, this is boring stuff and seems like a beginner basic, but the truth is that this is one of the first places that your text analytic efforts go from having great potential to being an effort with abysmal value. Look for not only a sophisticated technology solution from your provider, but also look to experienced analysts that know the art of using these advanced techniques to get the most from your data. You want a solution that can handle acronyms, mask for personally identifiable information, eliminate profanity, correct for misspellings . . .just to name a few.
- Don’t overlook the importance of integrating structured and unstructured data together for rich insights. Most of us today would no longer say we have a data problem, but an insight problem. We know there is more we can derive from the existing information we have – to help identify issues and even better do root cause analysis. Don’t just look for the categories and themes that customers are talking about, but understand how KPIs for those categories differ. Keep a pulse on the overall customer temperature by tracking not only key customer metrics, but also text analytic sentiment. Dive down into specific issues by combining all of the structured information about a specific topic alongside the insights you can garner from the text analytics about that topic to create a “mission control” for monitoring the issue and doing root cause analysis.
- Use text analytics to find and alert you of “hot button” conversations. When we give customers an opportunity to tell their experience stories during the CX process, they can say the darndest things. They might even hit on one of your “hot button” issues in their open-ended responses and dialogue– such as threats of legal action, survey manipulation, defection, or even possible instances of regulatory noncompliance. Use text analytics to identify these issues by creating categories to find them and use case management functionality to trigger alerts so the right folks in your organization can address them right away.
- Let text analytics be the hero to integrated multiple sources of customer feedback through a common category set. This is the holy grail of CX – look across all of your sources of CX feedback to get a 360 view of the customer experience. Text analytics is the right way to start this endeavor. By using a common category set, we have a framework to compare and contrast what is being said for example in social media vs the customer survey. While in some cases these sources tend to show corroborating results in other cases we find that we learn different things from one source or another. For example, the breadth of topics discussed in survey verbatims is often less diverse than what is discussed in social online reviews. We learn about topics that might not come up in the context of the survey often because of the directive nature of survey questioning.
- Don’t Forget About Data Mining in the World of Unstructured Data: One of the most impactful and prevalent analysis that our data sciences team does for our clients today is driver analysis to understand what aspects of the experience drive key performance or financial metrics. What is often overlooked is the power that unstructured data can have in that analysis. Often unstructured data will provide new insights as to critical pieces of the experience that drive outcomes. And when used in conjunction with a data mining tool such as the MaritzCX Spotlight tool, we can begin to understand the complex interactions and patterns of how those key experience aspects relate to demographic groups, psychographics, and more.
- Use Text Analytics to help eliminate the noise in your CX data for more focused analysis and insights. Let’s face it, many customer conversations are not between the customer and the brand. They are one to many in the social landscape. The trouble is that finding those nuggets in the deluge of what many call the “digital exhaust” can be overwhelming. The best in today’s text analytics can help us remove non-useful unstructured data from those customer experience gems. Using a technology which is very similar to the Naïve Bayesian Classifier approach which is used to remove spam from e-mail, we can train text analytics to pull out only the useful customer experience comments. This helps us provide a “CX filter” of the onslaught of social data and helps us get focused on the insights that matter.
So, as you think of your CX program and unstructured data, consider where you are in your evolution and adoption of Text Analytics. If you don’t have text analytics in your program yet, it’s time to dive in. If you do, consider how you can take that effort to the next level. Unstructured data is a beast that is going to continue to grow. The sophistication of our technology to handle those conversations – such as text analytics – need to keep pace or we will find ourselves missing critical insights.
This is a recap from a session Michelle gave at CXFusion 2016. Join us next year and learn more.