We've found that our clients struggle to identify trends in customer feedback across their many channels. Customer surveys, in-product feedback, support tickets, sales notes, and other inputs all contain valuable information, but analyzing all that data in aggregate is time consuming. How can you quickly track customer experience and continue to adapt to evolving customer needs?

Thanks to our expertise in AI, large language models (LLMs) and our years of experience collecting and analyzing customer feedback, we've developed a tool, Customer Feedback Trends, that simplifies this process. It automatically generates consistent, precise and insightful labels for vast amounts of open-ended customer feedback data.

To demonstrate how our tool works, we've fed it 500 Google Play Store customer reviews of a todo list app (source data available here). You can play with the interactive widget below to see the real customer feedback submissions people had and the auto-generated issue labels our tool created for them:

Our tool generates standardized labels and applies them to large volumes of open-ended text, creating data that can be easily counted and analyzed. Once text is converted into counts, you can quickly generate insightful graphs and charts. For instance, here is the breakdown of high-level feedback categories and the top issues raised by customers of the todo app:
Knowing the prevalence of customer issues enables critical insights. A quick look at the charts above generated for the todo app data reveals that many customers are dissatisfied with the app's pricing and value. Specifically, when you look at the top issues, nearly 10% of customers express dissatisfaction with the absence of a free reminders feature. With this new understanding, a marketer would have a good reason to experiment with the pricing of the reminders feature, possibly offering it for free to attract more customers.

Our tool also shows that 20% of the feedback is about feature requests, while another 11.8% is about bugs in the app. These are the kinds of issues that product managers and engineers often get asked to address. But without knowing how prevalent every customer request is, the issues sometimes get continually deprioritized. However, with our auto-generated issue categorization, we can quickly see the full list of feature requests and bugs and weigh each one:
With data like this, a product manager would have solid support to prioritize the "view completed tasks" feature before other requests. Similarly, engineers would know that investigating why customers are having so many sync issues would be the most impactful bug to work on.

Our tool doesn't just provide insights about customer issues. It also captures mentions of competitors and gauges the overall sentiment of each piece of feedback, enabling teams to spot trends in the competition and assess improvements in the customer experience over time:
Because this data is auto-generated, it's easy to run on a regular basis. By tracking these metrics over time, teams can gain more value as they assess whether their efforts to drive an improved customer experience are making a difference.

Linguists use the term "taxonomy" to describe the set of categories they use for labeling and understanding language. In our experience, manually labeling open-ended customer feedback in an accurate and actionable way can easily take a team of people weeks of time, depending on the volume of feedback. Additionally, building a taxonomy that not only fully covers the range of issues that customers have, but also has categories that are mutually exclusive from each other is an ongoing challenge.

Luckily, our tool auto-generates a taxonomy without human involvement, saving users considerable time and headache. For the todo app feedback, the auto-generated taxonomy contains over 200 distinct issues, but thanks to its hierarchical structure, it's relatively easy to follow. Click into the categories below to explore the todo app labeling taxonomy more:

Of course, clients may want to use their own existing taxonomies instead of generating a new one and our tool can accommodate this too. Even so, we've found the worst part of building a taxonomy is that you never really know if you did a good job! To address this, we've also built in measurement checks to evaluate the effectiveness of the taxonomy and suggest tweaks, giving users confidence in their taxonomy's efficacy.

Reach out with your customer feedback use case and we’d be happy to chat about what would work best for your organization. As a trial, we'll categorize your first 500 pieces of data for free.

We're available to help with a range of options from simply making our tool available to you for manual use, making an API available for programmatic use, or even helping you design and build something similar within your own organization. Additionally, we also can run qualitative interviews to explore customer feedback in greater depth and even help organizations build lasting voice of customer programs.

Customer Science Collective is a user research and product strategy consultancy with expertise in building generative AI products. We love the art and science of analyzing and acting upon customer feedback. We look forward to hearing from you!