Text analytics certainly has left me scratching my head a few times. Aside from figuring out what it even is, learning how to work with it took excessive time, patience, and – let's be honest here – more than a few sighs (yells?) of frustration. Although I'm constantly learning more, I have discovered some helpful hints along the way, and would love to spare you some of the head scratching I initially experienced. Try out some of these tips below and let me know what you think.
Narrow Your Focus
I'd argue that the planning stage is the most important part of TA. Jumping in without focus will lead you down paths of confusion and, quite possibly, rage… Yes, it can get that serious. Avoid roadblocks and popped blood vessels by answering these questions up front to help guide your project's purpose:
- What about my business do I want to improve, confirm, or find out?
- What data can I use to help me do the above? (Surveys, social media, case logs, emails, product reviews, etc.)
Draft a Dream Team
There are many moving pieces involved in a TA implementation, so determining who does what helps reduce confusion, overlap, or dropped tasks.
- Who are the key analysts, managers, decision-makers, and IT people from your company?
- How will they be involved in planning, setup, reporting, and analysis?
- Who's responsible for reviewing the data and making actionable recommendations?
Envision Your Finish Line
Ask yourself a couple more defining questions:
- What output do I expect to see, and how will it be used? (This is where we partner with Clarabridge, who helps automate the processing of unstructured data.)
- Who will use this data to improve, confirm, or dig deeper?
These questions help narrow your focus, but at the same time ensure that your efforts will produce something worthwhile and actionable. As mentioned in 3 Tips on Using Text Analytics to Capitalize on Untapped Customer Feedback, you ultimately want to "turn anecdotes into evidence that can drive action".
If you're still scratching your head (or, worse, if you've migrated to pulling out hairs) over text analytics, check out our real-life implementation example here, or leave a comment to explain your troubles. What would you like to learn more about? Where have you struggled or succeeded?