# Women in transition — Social Sales Safari protocol

Use when Margarita wants a detailed, source-grounded picture of women whose successful former identity has become too small, especially when she explicitly prefers public social research over interviews.

## Research boundary

Public social research is strong for:

- spontaneous vocabulary;
- transition triggers;
- push/pull/habit/anxiety;
- actions already taken;
- adjacent products and practices already bought;
- objections and dissatisfaction;
- self-reported geography and resource signals.

It does **not** establish willingness to buy an untested Cosmos Luminosity offer. Validate that through a real paid pilot, deposit, completion, outcome, and continuation/referral behaviour.

## Women-only inclusion rule

A record counts toward a women-only sample only when the author explicitly self-identifies through first-person public evidence such as:

- `woman` / `women` referring to herself;
- age-gender notation such as `34F`;
- explicit first-person bio/profile language.

Do not infer gender from name, avatar, voice, community theme, spouse gender, occupation, or writing style. Save the exact evidence in the ledger.

## Evidence ledger

For every record capture:

1. exact quote, preserving spelling;
2. direct public URL, not a search-result page;
3. date and platform;
4. explicit female evidence;
5. self-stated country/city/region, otherwise `unknown`;
6. explicit role and career/business stage;
7. resource signals: salary, capital, equity, runway, independent purchase authority — only when stated;
8. trigger;
9. push: what she wants to leave;
10. pull: what she wants instead;
11. what she refuses to lose or erase;
12. fears and doubts;
13. actions/alternatives already tried;
14. actual purchases and prices, if named;
15. what helped and what remained unresolved;
16. openness or resistance to cognitive, somatic, state/nervous-system, symbolic, spiritual, or energetic tools;
17. action readiness;
18. Cosmos Luminosity fit or exclusion.

Keep separate:

- ability to pay;
- prior category spend;
- urgency/action readiness;
- willingness to buy the proposed offer.

## Multi-source sampling

Seek several source contexts rather than merely several URLs:

- career/path and FIRE communities;
- founder/executive/independent-worker spaces;
- creator essays and newsletters;
- public professional posts;
- YouTube/podcast comments or transcripts;
- reviews/discussions of coaching, therapy, somatic work, retreats, and spiritual or energy practices.

If a source blocks access, make one serious attempt, record the bias, and switch. Do not spend the run repeatedly fighting one platform.

A platform-heavy corpus remains useful, but say exactly what it overrepresents. Example: FIRE communities illuminate capital, runway, salary, identity after work, and fear of leaving; they do not represent all successful women in transition.

## Geography reporting

Report three layers separately:

1. **Explicit counts:** only places authors named themselves.
2. **Unknown:** number/percentage with no usable location.
3. **Sampling bias:** where the platforms and search method over-sample.

Never turn `HCOL/VHCOL`, a remote job, currency, or travel pattern into a country guess. Never say “they mostly live in X” when X is only where accessible posts were found.

## Timeout-safe collection

Large 25-record fan-outs are vulnerable to timeouts and blocked sources. Use bounded passes:

1. Define one narrow cohort or evidence layer per worker.
2. Target 10–15 records per pass rather than 25 when source diversity is required.
3. Write a valid artifact after every five records.
4. Preserve partial artifacts on timeout.
5. Verify the shared filesystem artifact yourself; a worker summary is not evidence.
6. Audit record count, direct URL count, unique URLs/authors, platform mix, explicit gender evidence, explicit geography count, and purchase-price count.
7. If a worker timed out without a file, treat its findings as lost and do not cite them.
8. Continue with shorter gap-specific passes rather than rerunning the same oversized prompt.

## Confidence

- **High:** repeated unprompted theme across 5+ women or 3 source contexts, with behavioural support.
- **Medium:** 2–4 women or one source context with partial behavioural support.
- **Low:** one author, prompted answer, unclear profile, intention without behaviour.

Keep disconfirming evidence. Common useful disconfirmers include:

- she wants a lower-stress version of work, not retirement;
- she wants tactical career placement, not identity work;
- acute clinical, legal, safety, or financial needs take precedence;
- she is spiritually open but not financially resourced;
- she has capital but rejects inner-work language;
- she consumes self-development but does not implement.

## Compressed platform-as-research route

When Margarita wants to move to platform creation immediately, combine research and build:

1. Publish one recognition route for the narrow transition situation.
2. Add a short check-in that captures exact open-text language, stage, prior attempts/purchases, tool openness, and next-step readiness.
3. Return a useful first result — not only a thank-you page.
4. Offer a bounded paid next step or deposit.
5. Track recognition, check-in completion, paid conversion, completion, experiment/action, outcome, continuation, and referral separately.
6. Revise platform copy and routing only from repeated evidence, not single quotations.

This removes a long standalone discovery phase without pretending that public posts prove product demand.