How to collect user feedback that changes the product: the 3-channel system I used as Greenely's only designer, with 2,000+ users in a Facebook group.
If you want to know how to collect user feedback that changes the product, not just a dashboard of scores, use three channels at once: a place where users talk to each other (a community group), a place where they talk to you (calls, support, the people who show up), and a place where they act without talking (product analytics). Then only act on problems that show up in at least two of the three. That is the system I worked with as the sole product designer at Greenely, a Swedish energy app, where we had a Facebook group of more than 2,000 members, weekly meetings with random users, and, yes, people who walked up to the office and knocked on the door to complain or ask for a feature.
This post is the use case: what we heard, what we did with it, and what you can copy if you are the only designer on your team.
I joined Greenely in 2019 as a part-time product designer, and I was the only one. Greenely helps households track electricity use, see prices and lower their bills, with machine learning doing the optimization behind the scenes. You can read the full story in my Greenely energy app case study.
Before a design team existed, the app was built from concept wireframes that PMs drew. Prototypes went out without testing, and the architecture could not scale. So the problem was never "we have no feedback". It was the opposite. Feedback arrived from everywhere, in every format, and most of it was loud.
That is the part the usual "10 ways to collect feedback" lists skip. Collecting is the easy half. Sorting signal from volume is the job.
Here is how the channels worked in practice, and what each one was good and bad at.
A Facebook group with 2,000+ members is a gift and a trap. A gift because people describe problems in their own words, and other users reply "same for me", which is free validation. A trap because the most active members are not your average user. They are the power users, the people with an EV, solar panels and a summer house.
The value was the language. The pain did not arrive as "we need a notification center". It arrived as: I have no idea what is going on and when, so I keep opening the app to figure it out from the data. That sentence is a better brief than any feature request.
We also held weekly meetings with random users, and some users came to the office in person. Direct conversation gives you the "why" behind a complaint. When someone says "I use Greenely with one or two other apps", you can ask which ones and for what. That is how a vague feeling turns into a list of missing integrations.
The risk here is recency. The person who knocked on the door this morning feels more important than the many people who quietly stopped opening the app and never told you why.
So we added Mixpanel product analytics across the app to track specific events and clicks. This is the channel that keeps the other two honest. If the group says "nobody can find device settings", the funnel either confirms it or it does not.
Nielsen Norman Group frames this as attitudinal research (what people say) versus behavioral research (what people do), and their guide to choosing UX research methods is still the clearest map of it. My rule from Greenely is simpler: say plus do beats either one alone.
| Channel | What it tells you | Where it misleads | How we used it at Greenely |
|---|---|---|---|
| Community group (2,000+ members) | Problems in users' own words, plus "me too" validation | Over-represents power users and the loudest voices | Source of pain points and the exact phrasing for problem statements |
| Weekly calls with random users | The "why" behind behavior and the workarounds people use | Small sample, depends on how you ask | Checked whether group complaints held for ordinary households |
| People at the door, support | Urgent pain and emotional intensity | Recency bias: the latest complaint feels biggest | Flagged issues to verify, never a roadmap on its own |
| Product analytics (Mixpanel) | What users actually do, event by event | Shows the what, never the why | Confirmed or killed a problem before we designed for it |
| Internal workshops | Business priorities and constraints | Easy to confuse team opinions with user needs | Turned verified problems into priorities and buckets |
When we lined up the group posts, the calls and the data, the same pain points kept coming back. These are the ones published in the case study:
Notice that none of these are solutions. That was deliberate. If you collect feature requests, you get a list of other apps' features. If you collect problems, you get a product.
The tempting move was to ship the most requested feature first. Instead, the feedback pointed at one root cause: the app's structure could not hold what users wanted. Integrations, a feed, multiple homes, smart charging. Bolt those onto the old architecture and every new feature makes the next one harder.
So we ran workshops where we put every element of the app, current and planned, on the table and grouped them into buckets. Then we checked those buckets against the data and the user conversations: where people spent time and what interested them most. The UI moved to card-based elements so new content types could slot in as the app grew, and I built a design system so every new integration looked and behaved like part of one product. I wrote about why that kind of system pays off early in my post on product designer skills that matter.
The result, as published in the case study: more than 15 new integrations added in an organized way (20+ in total), a notification feed, clearer price signals, a smart charging flow, reorganized settings and a design system. The case study also reports retention up more than 80%, with users checking the app 5 to 10 times a week, households growing from under 20k to more than 100k in under two years, and 70% more devices added compared to the same period the year before. That was a team effort across product, engineering and leadership, not a feedback trick. But the feedback system is why we built the right things.
The same pattern held at much bigger scale. At ViewStats, the creator platform built by the MrBeast team, we had a Discord community of 17,000+ creators, plus interviews, onboarding analytics and support requests. The community told us creators were confused at the channel connection step. The funnel showed where they dropped. Together they led to the redesign I describe in onboarding UX that took activation from 13% to 65%.
And when you cannot reach users at all, like at Accern where the customers were large finance enterprises, you borrow the channel: we worked through the customer success team. That story is in my enterprise UX design post.
These days I also prototype fast to test feedback instead of debating it. A working build, like the ones in my AI design experiments, gets a more honest reaction than a static mockup, because people react to how it moves, sounds and responds, not just how it looks.
Where this does not apply: very early products with no users yet. There, you are doing discovery interviews, not feedback collection, and analytics will be mostly empty.
Combine one channel where users speak (calls, community, support) with one where they act (product analytics). Neither is enough alone. Speaking tells you why, behavior tells you whether it is common.
Read the places users already talk: support tickets, community groups, app store reviews, sales and customer success notes. Then book short calls with random users and ask about their last real session. At Greenely, a 2,000+ member Facebook group did more than any survey could.
I do not use a fixed number. I use a rule: the same problem shows up in at least two different channels, for example the community and the analytics funnel. That filters out the loudest voice and the latest complaint.
Rarely as asked. Treat requests as evidence of a problem, then design the solution yourself. At Greenely, many separate requests pointed to one fix: an architecture that could hold integrations, a feed and multiple homes.
If you are the only designer on your team and drowning in feedback, I am happy to help you set up the system and turn it into a roadmap. Work with me as your product designer, or browse my product design case studies first.
If you have a web or mobile app idea in mind or you need some advice about product design, feel free to get in touch ✉️