Community Guides

Operational detail for Diamond 1: choosing where to look

These are the steps behind Choose Where to Look. The chapters carry the judgment: why you are looking away from the center of a problem, what makes a community coherent, why an easy yes is suspicious. This page carries the procedure, in the order a diamond runs it: widen, narrow, then test before you commit.

The same steps in machine-readable form, for working with an AI, are Stages 1 through 3 of The Method Layer.

Diverge — Cast a Wide Net	A timed sprint producing 10 to 30 candidate communities with access signals.
Converge — Commit to a Community	Compare, notice the pull, assess access, select, and refrigerate the rest.
Test — Test Your Access	Confirm you can reach them repeatedly before investing in conversations.
Community Choice Gallery	Badly and well scoped communities side by side, with what makes the difference.

A community is not yet your people

What you choose here is a search space, not a customer. You are picking a population coherent enough that pains repeat inside it and varied enough that you can tell what causes them. Who your people actually are gets settled in Diamond 2, by who turns out to carry the validated pain, and the answer can land outside the community you started in. The procedure for drawing that boundary is From Community to People at the end of the pain guides.

Diverge — Cast a Wide Net

A 45 to 90 minute sprint that produces a shortlist of 10 to 30 plausible communities, each with a one-line profile and at least one access channel to verify later. Works solo; halve the timings.

You need sticky notes or a shared board, a timer, and two to five people if you have them.

Step 1 — Anchor a problem space (5–10 min)

Write one plain-language statement.

We are exploring friction around ______ for ______ in ______ context.

“Safety during the walking segments of commutes for women in urban areas.”
“Care coordination for adult children supporting aging parents across states.”

Solution words in the anchor

“App,” “platform,” “AI.” Any of these in your anchor has already decided the answer, and every group you generate afterward will be a group who might use that thing rather than a group who has that difficulty.

Step 2 — Generate communities (10–20 min)

Silent brainstorm for three to five minutes, then share round-robin. Aim for quantity: twenty or more. Stretch prompts when you stall:

Who experiences this most intensely?
Who is ignored by what already exists?
Who has workarounds nobody sees?
Who faces this at odd times or places: night shift, rural routes, transfer points?
Who is adjacent: caregivers, gatekeepers, frontline workers, volunteers?

Step 3 — Add texture (8–15 min)

For each promising group, write a one-line profile: role, situation, friction.

[Group] who [do X or live in Y] often [friction] because [why].

“Professional women who walk from subway to office often avoid side streets because lighting is inconsistent and foot traffic thins.”
“Night-shift nurses crossing hospital lots carry keys visible because patrols are sparse at 3am.”

Step 4 — Map an orbit (6–10 min)

Name one to three orbit roles per group: the people who influence or constrain the experience. Property managers, transit security, roommates, HR, campus police, rideshare drivers, forum moderators.

Orbit roles are access multipliers, and they reveal constraints you would otherwise discover late.

Step 5 — Capture light access signals (8–12 min)

For each group, note at least one channel where you could plausibly reach them within a week.

Physical — specific stations, entrances, bus stops, gyms, dorms, employer lobbies.
Digital — a subreddit, a Slack or Discord, alumni groups, Nextdoor, Meetup, congregation lists.
Warm routes — student clubs, coworking spaces, HR newsletters, union or local boards.

If you cannot imagine where they are and how you would approach, mark access LOW. Signals only. You are not concluding anything about access yet.

Step 6 — Snapshot and shortlist (5–10 min)

Build a one-screen table you can return to when you converge.

Group label	One-line profile	Orbit roles	Likely access channels	Access	Why this group might hurt
Women walking part of a commute	Subway to office, avoid side streets after dark	Roommates, property manager, transit security	Station exits, coworking list, local subreddit	MED	Lighting, sparsity, vigilance load
Night-shift nurses	3am lot crossing with fatigue	Hospital security, supervisors	Staff Slack, union board	LOW	Distance, staffing patterns
Female grad students	Campus to housing, late	Roommates, campus police	Grad Slack, department admins	HIGH	Staggered hours, low-cost routes

Keep ten to thirty rows.

Bias check (2–4 min)

Are most of these groups people like you, or people who are easy to reach? Add three that are neither.
Is at least one non-obvious: transfers, caregivers, gig workers, the person who cleans up afterward?

Diverge is complete when

You have 10 to 30 candidate communities, each with a one-line profile.
Each has at least one realistic access channel, even a tentative one.
Orbit roles are noted for your top eight to ten.

Field sheet

Anchor: We are exploring friction around ______ for ______ in ______.

Candidate groups (aim for 20+):
1) ______   2) ______   3) ______   ...

One-line profiles (pick 10-30 to flesh out):
- [Group] who [do/live] often [friction] because [why].

Orbit roles (1-3 each):
- [Group]: ______

Access channels (at least one per group):
- [Group]: ______

Bias check additions (3 non-obvious):
- ______   - ______   - ______

Into your data room: the dated sprint document, the shortlist table as a sheet, and any contact lists or channel links you surfaced.

Converge — Commit to a Community

Use this after the divergence sprint, alongside Commit to a Community. Convergence runs in four moves, and the output is one group plus a written record of what you set aside.

1. Compare. Lay the candidates side by side and record what you notice across them: size and coherence, evidence of neglect, whether they already spend money or time on workarounds, the access signal from Step 5. Write down the patterns, not just the ratings — a pattern across candidates is often more informative than any single row.

2. Empathize. Name the one to three groups that stay in your mind when you are not working. Write down why they feel different. This is evidence about whether you will still be curious in week four, and it is not a tiebreaker to apply at the end. Record it before you compare access, so it does not get rationalized.

3. Assess access. For each finalist, note what you actually know about whether they are observable, reachable, and willing to engage, and mark which of those three you are assuming rather than evidencing. Note your sampling risk: if everyone you can imagine reaching came through one door, your access is narrower than it looks.

4. Select. Name the group. Write the rationale in three or four sentences, and write the caveats and working assumptions beside it. Then write every rejected candidate into a durable note with the reason it was set aside. They are refrigerated, not discarded, and a failed access test sends you back here rather than to the beginning.

Reflection worth recording: what surprised you during convergence, which assumptions you updated, and what the next step is now.

Into your data room: the comparison board, the empathy notes, the access tracker, and the refrigerated list with reasons.

Test — Test Your Access

Use alongside Test Your Access. Before you invest in conversations or observations, confirm you can reach these people repeatedly. A few days here prevents wasted months.

Minimum access conditions

Count each only where evidence supports it, not where you have an assumption:

You can find them — specific places, online or offline.
You can engage them — a message, an intercept, an observation, an introduction.
They respond and open up — willing to talk, reflect, and tell you something you did not ask for.

Design the test

Record before you run it: which channels you are testing, how many people you will attempt, the method (message, intercept, post, visit), and the timeline. Draft the actual opening message, short enough to answer on a phone.

Keep score while you run it

What did you actually do?
How many people did you attempt to reach?
How many responded, and how long were the answers?
Surface or depth?
What barriers or surprises appeared?

A test you cannot summarize in numbers is an impression, and impressions are generous to whatever you already believed.

Read the signal

Green — strong access. People responded quickly and openly, and you could do it again next week.
Yellow — weak or no signal. Responses were few, shallow, or hard to replicate. Often this means too few attempts to mean anything, which is not the same as failure.
Red — blocked. You could not reliably find, reach, or engage them.

Then check the easy yes: if everyone who answered was already known to you, the test measured your personal network rather than your access to a community.

Update what you know

What do you now know about access to this group? Which assumptions were validated or overturned? Did you learn anything about subgroups, gatekeepers, or hidden barriers?

Next steps. Green, plan your first exploratory conversations. Yellow, refine channels, messaging, or sampling and run it again. Red, return to convergence and your refrigerated list.

Into your data room: the messages or posts you sent, the intercept script, and the record of replies including the silence.