How to Run User Research in a Discord Community With a Standing Panel
Open calls for volunteers return the same voices every time. A selected panel of thirty to sixty members, asked on a fixed rhythm, produces findings a product team can act on.
What the open call actually returns
A product manager posts asking for input. Nine replies arrive quickly, six from the members who reply to everything. Two calls, one no-show, one write-up beginning with the word members.
It is quick and it feels like listening. What it lacks is a denominator, so nobody can distinguish a market signal from an enthusiastic minority with spare time.
The compounding cost is worse than the immediate one. Open calls teach the enthusiasts to expect access and teach everybody else that research requests are not for them, so each round of input is narrower than the last.
What a panel returns instead
Forty two names in a document, each tagged with segment, tenure and usage band. Eleven get selected because they match the decision. Three questions go out. Twenty eight responses arrive inside two days, because agreement to be asked was secured months ago.
Same effort. The selection simply happened in advance rather than being handed to whoever happened to be online.
Composition
Thirty to sixty members. Fewer than thirty makes segmentation impossible, more than sixty makes the relationships impossible, and the relationships are what hold the response rate up.
| Axis | What to include | What that group uniquely sees |
|---|---|---|
| Tenure | First month, first year, long standing | New members can still see the onboarding |
| Segment | All customer types in rough proportion | Prevents a roadmap for one audience |
| Usage | Heavy, moderate and light | Light users know where the product lost them |
| Sentiment | Two or three specific complainers | They have already done the analysis unpaid |
Recruit privately, one message at a time, stating the frequency and the length of commitment. A panel recruited through a public announcement is an open call wearing a different name.
Rhythm
Hold one cadence for two quarters before adjusting it. A workable shape: a three question set every two weeks that closes in forty eight hours, two half hour calls a month, one themed open thread per quarter, and one quarterly note back to the panel describing what changed.
That final item is the one most teams skip and the one that decides whether the panel still exists next year. People participate to see their input become something real, and silence reads as extraction.
Compensation that selects for the right people
Cash works and quietly recruits people motivated by the cash. Access recruits people motivated by the product.
What holds up over time: genuinely early looks before things are polished, a direct line to someone who can answer rather than log, a named credit on a shipped change with permission, a visible community role that marks contribution, and a modest gift for calls that respects the time without becoming the reason for it.
Recognition is the most durable of these. A one-time payment is spent. A role and a credit stay visible.
Making findings travel
Research that never leaves the community team is a hobby. The panel earns its budget through a document other departments read.
One page, every month, identical structure: the question, the number of respondents, three lines of what they said, and the decision it informs with an owner beside it. Circulate it to product, support and marketing.
Writing the decision next to the finding turns the report into a record. When someone eventually asks what the community programme produced, you can trace a shipped change to a specific response on a specific date, which ends the conversation immediately.
Findings without decisions attached are anecdotes with better formatting.
The two failure modes
Capture happens when the same eleven names get every request because they answer fastest. Track who has been asked in the last quarter and rotate deliberately.
Fatigue arrives through frequency creep. Launch pressure pushes the cadence to weekly, habit keeps it there, response rates slide, the fastest respondents dominate again, and the open call has been reconstructed inside the panel.
The surrounding layers
A panel is one component. It does not summarise what the entire membership is saying, which needs a separate reporting loop, and it does not catch problems arriving through support, which needs routing and escalation. It also increases moderation load, since a more engaged membership produces more of everything.
What it removes is the habit of calling guesswork listening. A selected group, a steady rhythm, and decisions written down beside findings turn community input into something a roadmap can be built on.
Ask a chosen group on a schedule and write down what you did about it. That is the whole method. More at danieljeong.org.
Want a properly managed community?
Get Your Free Analysis →