Discovery and Exploration

What’s in a bad user review? Maybe your next breakthrough

Share this post:

Key Points:
– Not all reviews are created equal. In fact, some reviews, even negative feedback, can reveal insights and better customer experiences.
– Manually reading, categorizing and prioritizing customer feedback can be time-consuming and costly. Fortunately, IBM Watson doesn’t have these constraints.
– Use Watson Discovery Service to analyze reviews with three stars or less and positive sentiment to uncover golden nuggets of actionable insights buried in the feedback.

Learn more about Discovery Service


Think you know your customers and what they want from your app? You might be surprised. Gain new insights from customer reviews.

There’s no question that customer reviews are valuable. Especially when it comes to mobile applications, where one review with less than three stars can dramatically reduce the number of downloads and stall adoption.

But, not all reviews are created equal. Some negative reviews can reveal insights that boost your app’s popularity or at least deliver a great experience for your customers.

While star ratings provide an overall score for an application, it isn’t enough. The trick is distilling meaningful feedback from reviews including both positive and negative feedback. In IBM’s work with thousands of clients, we’ve found that regular app users can be the harshest critics, and one of the best sources for ways to improve the customer experience.

Taking a closer look at customer reviews

Imagine a fan goes to see their favorite band in concert but leaves disappointed because the venue had poor acoustics. Now that fan might give the band a three-star rating along with this review: “I love this band but the sound was too loud!”

We see the fan uses the word “love” and this provides a clue. In IBM Watson terms, this is called “positive sentiment” and it’s a powerful concept. Positive sentiment can be a great indicator of which reviews could contain useful and actionable information. In this case, it’s easy to identify the problem, which was a bad venue. The band can quickly remedy the situation by selecting new venues with better sound and improve the fan experience.

Conversely, reviews like “I hate all music by this band” are judged to have negative sentiment and are less valuable. While bands want everyone to love their music, it rarely happens. In this case, there isn’t much that can be done to turn around a concert-goer with such strong negative feelings.

The concert-goer scenario is just one example of how developers can use Discovery Service to uncover a goldmine of ideas for improvements that will delight users. Similar nuggets can be found in almost any customer feedback or product reviews you can imagine, for example:

  • “I love this book but I wasn’t able to relate to the new character introduced into the series.”
  • “This is my favorite restaurant for family celebrations. The last time we went, the signature dish was no longer on the menu.”
  • “I love my new car but the driver-side cup holder is too far back and out of reach.”

Of course, evaluating a few reviews manually doesn’t take long but what if you have thousands of customer reviews? Reading, categorizing and prioritizing them all can be time-consuming and costly. Fortunately, Discovery can automate this entire process.

How it all works

Our Mobile Innovations team used Discovery to build a simple, sample iOS app to evaluate recent reviews of the top 100 free apps in the Apple app store.. As part of the analysis, we directed the service to calculate and display three insights for each of the reviewed apps, which developers could then use to identify and prioritize next steps:

  1. Sentiment – The overall sentiment of all reviews for the app, which is measured by assigning a value from one to ten to each review and then averaging the total. Then, the app is assigned a letter grade from A to F based on the overall average sentiment score. The higher the grade, the more positive reviews for that app; the lower the grade, the more negative reviews.
  2. Keywords – A list of the most common words found in an app’s reviews. Keywords often map to a feature or concept in the app including how often each keyword is mentioned in positive, negative or neutral reviews, which makes it easy to identify popular or polarizing app features.
  3. Opportunities – A list of three-star or lower reviews that also include positive sentiment, which an app developer can then analyze to discover opportunities for improving their app based on customer feedback.

See for yourself the types of insights that had been hidden in the reviews and which became visible with Discovery.


IBM is making it easier to analyze relevant customer reviews and uncover actionable intelligence buried in the data. Use this tutorial and get the code to create your own instance.

Learn more about IBM Watson Discovery Service

Explore how Watson Discovery Service can help engage your customers

Offering Manager in the Mobile Innovation Lab, IBM Cloud

More Discovery and Exploration stories
July 16, 2018

Building trust in the age of AI – How businesses can build fairness into their machine learning models

Every organization that develops or uses AI, or hosts or processes data, must do so in ways that allow them to rationalize the decisions or recommendations in a way that is easily consumable. Let's examine Forrester's recommendations how organizations can leverage AI for the good of humankind, while avoiding the ethical pitfalls associated with perceived discrimination.

Continue reading

June 26, 2018

Updates to Watson Visual Recognition – 
Price reduction for Custom Classification, and 
Food and Explicit Models are now GA

Announcing updates to the IBM Watson Visual Recognition: A price reduction for Custom Classification events, and two models becoming generally available. Our blog highlights all these exciting changes.

Continue reading

June 25, 2018

Faster data discovery and access – Forrester Names IBM a Leader in Machine Learning Data Catalogs

IBM Watson Knowledge Catalog was recently recognized by Forrester as a Leader in: “The Forrester Wave™: Machine Learning Data Catalogs, Q2 2018.” IBM’s MLDC offering, Watson Knowledge Catalog, is a powerful tool that developers and data scientists can use to index all the available data sets in their business.

Continue reading