How to Fix Slow Response Times in a Discord Community When the Answer Lives in Another Team

Start by finding out who is actually late

Take the last few dozen questions that went unanswered for more than a day and put each one into one of two groups.

1. Questions somebody in the server could have answered from what they already knew.
2. Questions that required a person outside the server to open a system and check something.

Group one responds to staffing. Add hours, add people, add alerts, and the wait shrinks. Group two ignores staffing entirely. You can double the moderation team and every one of those questions will wait exactly as long as it did before, because the answer was never in the room.

That second group is where most of the visible damage comes from. It contains order status, refunds, account access, shipping problems and release dates, which are the questions members care about most and are least willing to wait on.

One number is hiding the problem

Most communities report an average response time. It merges two waits that belong to two different owners, so it can never tell you which one to fix.

Use two:

Once these are separate, arguments about performance become short. Fast acknowledgement and slow resolution is a handoff problem. Slow acknowledgement is a coverage problem. You can act on either, which was never true of the blended figure.

What breaks at each size

The same broken handoff shows up differently depending on how big the community is, and the right fix at one size is over engineering at the size below it.

Around 100 members. One person handles everything and asks colleagues directly when they get stuck. This works. The only thing to do here is keep a running note of which questions had to leave the server and who answered them.

Around 1,000 members. Asking for help turns from an occasional favour into a daily stream, and it lands in the personal messages of people who never signed up for it. Answers arrive whenever those people have a gap. Move the requests into one shared place and agree, in writing, who covers what.

Around 10,000 members. Informal agreements stop holding. Each category of outside question needs one named owner and a turnaround that owner agreed to. The community team must also be allowed to repeat that turnaround to the member, because a member who is told when to expect an answer waits, and a member told nothing assumes they were ignored.

Around 100,000 members and beyond. A chat message asking a busy colleague for help cannot compete with the work already sitting in their own queue. Escalations have to be created as records inside the owning team's system, carrying the member's context, with a status the community team can read without chasing anybody.

The pattern to watch for is simple. Every time the community doubles, the informal part of your process is the part that fails first.

Write the agreement down in one place

Whatever size you are, the artifact is the same and it fits on one screen. For each type of question that leaves the server, record the owning role, how quickly it comes back, and what the member gets told while they wait.

The fourth column is the one teams skip and it is the one members feel. A stated wait is tolerable. An unexplained silence is not. Use the role name rather than a person's name so the agreement survives someone changing jobs.

Run the audit for one week

Before any of this gets built, get the evidence. For seven days, log every question that did not close the same day and write next to it the team that would have had to answer it. Note whether the member heard from a human while they waited. At the end of the week, group the log by team.

The result is a short list of the departments your members are waiting on, ranked by how often. It takes almost no effort, it is impossible to argue with, and it is a far better basis for a leadership conversation than a request for more moderation hours. In most cases it also quietly clears the community team of a problem they were being blamed for.

The layers around this one

Escalation is one part of response time, not all of it. Coverage hours decide whether anyone is present when the question arrives. Channel and role structure decides whether it lands somewhere anybody is looking. Documentation removes the questions that never needed a person. Automation handles instant acknowledgement. Reporting makes the whole thing visible to the people paying for it.

Fixing escalation stops questions from vanishing into other departments. It does not build the rest, and knowing which of those layers you have is its own useful audit.

Members rarely remember how fast an answer came. They remember whether anybody ever came back to them at all. Read more at danieljeong.org.