The one-participant check

The one-participant check asks one question of any result: how much of it rests on a single person? You work the result out with everyone in, then again without the person who moved it most, and print both numbers side by side.

What did the Beyond Differences team find?

Five students in Karachi wanted to know whether young people from different community backgrounds felt more comfortable talking about ethnicity across those lines after a facilitated dialogue. They ran it twice, with two different groups, and measured comfort before and after on a five-point scale.

In the first session, average comfort rose by 1.14 points, from 3.29 to 4.43, among seven people. In the second, run with a tighter method and an outside facilitator, it rose by 0.43 points, from 3.71 to 4.14, among seven people.

The result got smaller every time the method got better.

Then the programme’s weekly review spotted what was driving the second number: one participant, who accounted for most of the movement. The review asked the team to show it. They did. They kept that participant in the headline, and printed the check beside it: take that one response out and the second session’s result is a fall of 0.17.

They also kept in the data the two participants whose scores went down.

Figure	Average comfort change	Detail
Session 1	+1.14	3.29 to 4.43 on a five-point scale, seven people. Range +0.86 to +1.14, because five participant codes had to be re-matched after the event.
Session 2	+0.43	3.71 to 4.14, seven people, a different group, tighter method, outside facilitator.
Session 2 without the largest mover	−0.17	The same session, with the one participant who moved most taken out.

Why is the smaller number the better one?

A smaller number isn’t a worse result. It’s a better instrument.

A first measurement uses the first method you thought of, and that method has slack in it. Tightening it takes the slack out. What is left is the part you can defend.

How do you run the one-participant check?

Four steps, and no software.

1. Work out your headline result with everyone included.
2. Take out the participant who moved it most and work it out again.
3. Print both numbers side by side, labelled, in the same size of type.
4. Keep that participant in the headline. The second number is a check, not a correction.

To go further, take each person out in turn and print the lowest and the highest.

If the two numbers are close, no single person is carrying the result. If they are far apart, as in Karachi, say so. Either answer is a finding.

The check matters most in small groups. In a team of five, or a session of seven, one person is a large share of the room.

What does the check not tell you?

It does not tell you why a result moved, or whether the people in the room were typical of the people outside it. The Beyond Differences scores are small, self-reported and have no comparison group, and the two sessions were different groups of people. The check tells you how fragile a number is. That is enough to keep a team honest.

Where does this come from?

Beyond Differences ran this inside the Grassroots Internships 2026, delivered with Peace Through Prosperity. Teams were scored against a published scoring guide that gives thirty of its hundred marks to how honestly a team handles its own evidence. Peace Through Prosperity calls this finding the most transferable thing the programme produced.

Read what happens next in the Grassroots Internships working guide, watch the teams present on the Grassroots Internships page, or run an experiment of your own from The Grassroots Playbook.

So here is my question, and I would like an answer rather than agreement: the last time you reported a result, did the smaller number sit beside the bigger one?

If you run a Playbook experiment and print both numbers, send them to me. I would like to see them. Get in touch.

Kubair Shirazee wrote the Grassroots Playbook, produced in partnership with Peace Through Prosperity. He works on how decisions get made, in enterprises, communities and the spaces between.