Overview
How you structure your teams in Relay directly affects the signal you get. This guide covers the right size for a check-in group, when to split a team into separate groups, and what breaks when you get the structure wrong.
Start with the working unit, not the org chart
The most common mistake when setting up Relay is mirroring the org chart. You have five departments, so you create five teams. Each team has 40 people. The results come back and tell you almost nothing useful.
Relay works best when a team reflects a real working unit: people who share a manager, work on the same projects, and have a common day-to-day experience. A group of 40 people with several different managers and offices have almost nothing in common when it comes to their weekly experience. The signal you get from asking them collectively is an average of several different realities. That's not insight. It's noise.
Think about the person who would read and act on the results. In most cases, that's a direct manager or team lead. Structure the group so the results land with the person who can actually do something about them.
How big is too big
Beyond 20 to 25 people, check-in scores start to flatten. High and low sentiment cancel each other out, trends become harder to spot, and the result feels generic. You end up with a number that could describe the experience of almost anyone in the group.
More importantly, a large group changes how honestly people respond. Individual responses feel invisible when 40 people are in the group. People sense that what they say won't move the aggregate much. That feeling reduces candour. The check-ins still come back, but they come back blander.
If you're working with a large team, split before you start. Two groups of 12 will give you more useful signal than one group of 24, and you can spot differences between them. If the two groups are having very different experiences, you want to know that.
How small is too small
Relay requires a minimum threshold of responses before a score appears. This protects anonymity: if a team has three people and two answered, one person can deduce that the scores came from the other two. The threshold means this doesn't happen.
Practically, teams of fewer than five members will often not generate enough responses to see results consistently. It doesn't mean you shouldn't include them. It means you should be realistic about what you'll get and consider whether they'd be better folded into a slightly larger group that shares enough working context.
Four people who work closely together and share a manager are worth including. Two people who work on entirely separate things are better served by a different structure.
When to add a second team
Add a second team when you have two groups whose day-to-day experience is genuinely different. A studio team and a client services team at the same agency. A shift team and a daytime team at the same facility. A regional office and head office. If the two groups have different working conditions, different pressures, and different managers, combining them will obscure both of their realities.
Don't add a second team just because you want more data. More teams means more Monday summaries, more check-ins to read, and more loops to close. Every team you create is a commitment to respond. Start with one or two. Add more when you have the capacity to act on what comes back.
The goal is not coverage. The goal is signal you can act on.
Set up fewer teams than you think you need. Get the structure right for those first, close the loop consistently, and let the results tell you whether another group makes sense. Good signal from two teams is worth more than approximate signal from ten.