BlogAll Text Analytics Systems are NOT the Same

All Text Analytics Systems are NOT the Same

Claiming that all Text Analytics solutions are the same is like saying all forms of transportation are the same. They’re not.

Research your Text Analytics options. Ask your vendor, even Mindshare, all of these important questions. You’ll be happy you did.

  • Does the vendor’s text analytics platform categorize comments with similar-themed keywords and phrases for comparing and trending? (For example, would the comments “lukewarm cheeseburger” and “the hamburger patty was cold” both be classified under “Hamburger” and “Temperature”?)  Or would you have to mentally group terms such as “cold” and “lukewarm” to understand the scope of a problem.
  • Does the vendor’s text analytics accurately identify the key topics in each comment individually as they come in? Or does it require a large sample size and batch process?
  • Does the vendor’s product ensure the highest quality results by using custom-tuned Natural Language Processing (NLP) semantic rules? Or is it based on simple keyword extraction or statistical probabilities?
  • Is their insight tagging at least 90% accurate overall? Or is it only slightly better than flipping a coin?
  • Does the vendor’s solution minimize setup time with pre-tuned, industry-specific models? Or is the vendor’s solution generic, lacking industry domain knowledge and requiring extensive time and effort?
  • Does their text analytics product seamlessly integrate with your current EFM system (Enterprise Feedback Management) or will you have to manage two separate systems? (E.g. Will you benefit from the cost savings of a fully integrated EFM and text analytics platform?)
  • Does the product automatically highlight correlations between structured survey data (like location, time of visit, and satisfaction ratings) and unstructured survey data (open-ended comments and social media)? For example, would your product be able to find that the phrase “on … cell phone” is highly correlated to poor customer service scores? Or will you have to connect the dots yourself?
  • Does their system send instant alerts when a key issue is detected in a customer comment? For example, if a customer reports “slipping and falling” on a wet floor, is an alert email sent immediately to your corporate office? Or will you find out much later?
  • Are their text analytics results reported in real-time and available 24/7 to managers at all levels of your organization (permission-based)? Or will the results be stale? Can the reports be pre-scheduled to email to managers or must you always log in every time to get what you need?
  • Do they have a turnkey product that can identify the central themes and sentiment in customer comments? For example, could it identify that the phrases “rude cashier” and “check engine light” are showing up frequently in negative context? Or do they require significant training, setup time, and resources to reach that insight?
  • Can the vendor customize their solution to tag comments in categories unique to your business, such as products, programs, competitors, etc.? Or is it limited to measuring generic insights only?
  • Can you navigate directly to source comments about key performance areas or are their results limited to high-level statistics? For example, assume that several feedback comments mention the cleanliness of floors in a store; can you instantly pull up those specific cleanliness comments to identify the root of the problem?
  • Do they support more than seven major languages? Or are they limited to English and not much else?
  • Is their solution fully up and running (already developed) and ready to deploy within just weeks, or are they selling something you’ll have to wait months and years to ever see.
  • Does a team of professional, full-time text analytics experts guide their text analytics solution? Or is their text analytics package just a piece of software with no support?
  • Are they confident enough to show you an unscripted live demo of their solution?  Will they allow you to test it with your own comments to see real results using a sample of your own data?  Or are they hiding behind a staged presentation?