How Analyzes NPS Responses with Machine Learning

How Analyzes NPS Responses with Machine Learning

This is a guest-post by Chad Keck, CEO & Co-Founder of, a customer success platform that allows companies to measure loyalty and customer sentiment using the Net Promoter Score (NPS) system.

Let’s play a game. Later in this post, I’m going to name three different machine learning service providers. Based on the one you pick, like a fortune teller, I’m going to tell you a little bit about yourself. Sound fun? Good. But first…

Machine Learning: the two most over-hyped words in today’s lexicon.

Seems like all machine learning is good for is asking Siri to play a song or self-driving cars that are far too expensive for me to afford. They have a saying here in Texas: all hat, no cattle.

But what if? What if machine learning could do useful things for me? Hard things? Things that actually saved me and my team time? Hours or even days of busy work? That adds up quickly.

What could machine learning do, though? That’s the big question. What is something that takes a lot of time, is very manual, and machines would be good at doing instead?

Analyzing Verbatim Feedback From NPS Surveys

Going through and tagging NPS results with trends/themes can be a manual and time consuming process. It’s also one of the most important differences between the companies getting the most out of NPS and the ones who just want the score to report to their boss.

Regular readers of this blog know that NPS can be used for so much more than just a score. NPS can be used for sales and to growth hack your business.

But I get it, who has time to slog through thousands of responses, tag them individually, and then review the graphs to get the insight needed to kick your NPS to the next level?

You do…

Introducing Machine Learning Powered Keywords

Important note: Keywords Analysis is NOT a replacement for tagging (how we power Trends Analysis), nor will it ever be. It’s a complementary form of analytics and does not auto-tag for you.

But if you don’t have the time to initially invest in manual analysis or just need a quick sanity check of top drivers of your customer sentiment, you can now access a one-click machine learning feature for NPS. I’m serious, one click, it’s that easy. Signup now and find out.

How We Did It (a.k.a. Time For The Game)

Finally, it’s time for the game! In order to build this killer feature, we had to evaluate a range of solutions. In the end, there were three clear leaders in the space:

  • IBM’s Watson (the one that won Jeopardy)
  • Amazon Comprehend
  • MonkeyLearn

They were all very good and comprehensive, but for our needs, one of these services stood out from the rest. The one you guess it is will tell me a lot about you.

If You Guessed IBM Watson …

… Then you’re the type of person who closely follows the tech news, but you’re probably once (or twice) removed from the everyday engineering… more of an executive type. CxO. Vice President.

You know that IBM Watson has its hands in everything from medical supplies to your taxes (thanks to H&R Block), but you haven’t looked under the hood to explore the APIs and evaluate the services for yourself.

If You Guessed Amazon Comprehend …

… Then you’re a builder. Someone who knows that Amazon Web Services are generally pretty terrific out of the box and have a lot of functionality that is a couple of years ahead of most other providers out there.

IBM Watson has made far more headlines, but for all the hype, the services are not really all there yet. It’s kind of a hodgepodge of APIs that don’t really work together well. IBM could well turn it around and take the lead, but right now they are flailing a little bit and letting competitors make up the lead.

If You Guessed MonkeyLearn …

… Then you should seriously come work for us, you’re clearly a genius. =) But really, most of you are probably like Monkey-What? Isn’t that a mailing list solution? (No. That’s MailChimp. Totally different.)

MonkeyLearn hits that sweet spot of being easy to use AND being extremely powerful.

Easy to use is a hard thing to pull off in machine learning land, trust us. And it’s also important because we want to be dead simple for you to use, and if our own tools are too complicated, they might end up confusing you.

The Power of Machine Learning In One Click

Too much of machine learning is either:

  • Too Dumb
  • Too Hard

We wanted to create the Goldilocks of machine learning. Something that:

  1. Solves a real problem
  2. Saves a LOT of time
  3. Takes one click

Did we succeed? There’s only one way to tell: Try It Today

This feature is in beta, and since Keywords Analysis is most useful for larger data sets, we’re only releasing this on our Grow plan and up (for now). So it’s not available to everyone, yet. But based on your feedback, we may evaluate expanding this to other account levels.

What do you think?

We are all using it here internally at for our own NPS surveys and it has surfaced some very actionable data from our feedback that wasn’t as obvious before. We hope it will do the same for you.

Originally posted in's blog.

Chad Keck

September 20th, 2018

Posts you might like...

MonkeyLearn Logo

Text Analysis with Machine Learning

Turn tweets, emails, documents, webpages and more into actionable data. Automate business processes and save hours of manual data processing.

Try MonkeyLearn
Clearbit LogoSegment LogoPubnub LogoProtagonist Logo