Sentiment analysis: Empower your team by sharing customer love

  • 24 November 2022
  • 3 replies


🤖 Often the feedback and requests product teams centralize in Productboard focus on the negative, like customer painpoints and shortcomings of today’s product. It means too often we miss out on hearing what customers love about our product!

Capturing customer love and sharing it with the team is a great way to celebrate successes and boost morale, identify your product’s value propositions in customers’ own words, and identify fans of your product who might provide a testimonial, take part in beta programs, or join your customer advisory board.

👉 With Productboard’s new sentiment analysis it’s easier than ever to surface and share this customer love with your team!

  • Productboard can now intelligently detect the sentiment in each feedback note
  • Filter by sentiment (positive, negative, neutral) to zero in on customer love (or else your customer’s greatest painpoints)
  • Combine your sentiment filter with other filters (such as product area, customer segment, or topic) and save it as a view. Then share with colleagues.

👋 Interested in trying out sentiment analysis? Productboard customers on the Scale and Enterprise plans can now sign up for the private beta.

3 replies

Userlevel 6
Badge +12

Oh wow. 
This is definitely going to be a really interesting feature as it matures. 😅 I’ve done something similar to this for another product and it can be a glorious feature - especially when it comes down to the value of identifying just how a user, customer, or company feels. Sounds like it’ll help with sorting out feedback based on sentiment and what type of project it came from. 

Great work, Productboard. Shoutout to the PMs and engineers working on this one, I know it’s not an easy feature to throw together.


Thanks David! Could you share what were the biggest challenges you had to face? 

Userlevel 6
Badge +12

Thanks David! Could you share what were the biggest challenges you had to face? 

When it comes to sentiment analysis there are a lot of “known” issues (like sarcasm, tone, polarity, and idioms). But the biggest issue that I ran into was providing meaningful insight for the right circumstance. Since we were a contact center solution there were a LOT of scenarios where the context could mean everything. 

We ended up taking a slightly manual way of assigning context (we broke it down from inbound vs outbound, queue the call came in on, and the caller’s known position in an organization). Many of this we captured automatically but there were times where manual input was required to ensure that it mapped correctly. This beta seems as though it’ll have some of the context-building scenarios through additional filters within Productboard - so this is your way of “manually assigning context.” 

My guess is that you’ll find that the algorithm isn’t flawless and you’ll have to continue to curate the content, even if you’re using a third-party system that’s well-known for its sentiment analysis (we used Amazon in our first iteration).Just make sure users fully understand that the algorithm will take some time to get used to how their users are being interpreted and that some manual evaluation may be beneficial.