# Research SOP

This SOP keeps CultureRadar manual, transparent, and falsifiable. The goal is not to score companies with false precision. The goal is to identify culture inflection signals and make timestamped, testable observations.

## Weekly Rhythm

Monday: alerts, transcripts, and red flags
- Review Google Alerts for all selected companies.
- Open the latest IR pages and earnings materials.
- Record only material events in the workbook.
- Ignore low-signal repeat coverage unless it changes the interpretation.

Tuesday: employee-review trend
- Check Glassdoor manually for each active research company.
- Record overall rating, CEO approval, business outlook, and management confidence where visible.
- Read 20-30 recent reviews when preparing a card.
- Capture theme drift, not isolated anecdotes.

Wednesday: talent flow
- Use LinkedIn manually for VP+, Staff+, Principal, core product, core engineering, and business-unit leadership movement.
- Record quarterly inflows, outflows, destination/source companies, and role seniority.
- Separate executive departures from broad workforce movement.
- Treat LinkedIn as a directional sample, not a complete census.

Thursday: selection and synthesis
- Pick the one strongest publishable judgment for the week.
- Do not force a judgment if the evidence is mixed or weak.
- Draft the card using the template.
- Lock the primary validation metric, threshold, and deadline before publication.

Friday: publication
- Publish one bilingual issue.
- Update the judgment tracker immediately after publication.
- Post one English X/FinTwit thread and one shorter Chinese thread.
- Log reader feedback and distribution notes.

## Evidence Standards

Every published card must include four evidence lines:

1. Talent flow
2. Employee review trend
3. Management language
4. External red flags

Each evidence line needs:

- A plain-English interpretation.
- At least one source link.
- A direction label: improving, stable, deteriorating, mixed, or unknown.
- A note on why the evidence matters.

## Judgment Rules

- A judgment can have many watch items but only one primary validation metric.
- The threshold and deadline are locked at publish time.
- Direction right but timing wrong is a miss, with a note.
- Acquisitions, delistings, major restatements, or clearly pre-defined exogenous shocks may be marked void.
- Partial outcomes must be rare and tightly justified.

## Company Baseline Completion Definition

A company baseline is complete when the workbook has:

- At least one talent-flow entry covering the latest 4-8 quarters or a documented reason why the sample is insufficient.
- A current employee-review snapshot and recent theme notes.
- Notes from the latest 2-3 earnings calls or official transcript equivalents.
- Red-flag search results or a clear "no material issue found" note.
- A draft direction field, even if the answer is "stable" or "no publishable signal."

## Reader Feedback Questions

Ask the same three questions every two weeks:

- What is the core claim you took away from this card?
- Which evidence line made the claim more or less credible?
- Would you keep reading this weekly if it remained free for now?

Do not lead with payment questions during the first three months.