Skip to content
Module
Module 3 of 6Lesson 1 of 3~10 min

Take stock of what your customers already tell you

Your support desk, app stores and surveys pile up feedback that nobody has time to read. This lesson helps you list these sources and understand what each can, or cannot, teach you about your customers. You will know where to start before handing any dataset over to a language model for analysis.

Updated November 14, 2026.

In short

Existing customer feedback covers support tickets, app store reviews, open answers from NPS surveys and notes from sales calls. It gives a large volume at low cost but mostly shows the what, seldom the why. For a PM, it is a way to pick the right interview topics rather than a substitute for talking to people.

Lesson objective

By the end of this lesson, you will be able to inventory your customer feedback sources and say, for each one, what it reveals and what it does not show.

Questions this lesson answers

  • Which customer feedback sources should you use in product discovery?

    The most common are support tickets, ratings and comments on app stores, open-ended answers to NPS surveys and written summaries of sales calls.

  • Are support tickets representative of your users?

    No: they mostly come from people who hit a problem and took the time to report it. They flag real difficulties without telling you how many users share them.

Topics covered

  • analyze customer feedback
  • support tickets and app reviews
  • NPS comments
  • customer feedback bias

Where it fits

Use what your customers already tell you

How do you get a reliable reading out of hundreds of tickets, reviews and comments, and know what it does not tell you?

Lessons in this module

  1. Take stock of what your customers already tell you (this lesson)
  2. Code feedback at scale with an LLM
  3. Check an LLM's coding on a sample

What you will learn in this module

This lesson is part of the course Set up continuous discovery with AI

  • Analyze existing feedback (tickets, reviews, NPS, calls) at scale with an LLM, using a taxonomy, structured outputs and a validation sample.