How we will measure a system without pretending a survey is a census.
Public trust depends on being precise about what the data can and cannot establish.
1. Separate official statistics from crowdsourced experience.
EEOC statistics are reported as agency statistics and linked to the original source. Contributor data is reported as responses from people who chose to participate in this project. We will not describe a self-selected sample as representative of all EEOC charging parties.
2. Standardize the core questions.
Every contributor is asked about dates, office, investigative contact, witness contact, document requests, mediation, status updates, outcome, and perceived impact. This lets us calculate measures consistently.
3. Protect small groups.
Office-level results should not be published where the sample is too small to be meaningful or could expose a contributor. The publication threshold should be set before public dashboards launch.
4. Distinguish “No” from “I don't know.”
For example, a charging party may not know whether the EEOC requested employer documents. The form preserves “Not sure” instead of forcing a negative answer.
5. Explain verification status.
Stories can be marked self-reported, partially documented, or documented. Verification means the project reviewed relevant publishable records; it does not mean the project adjudicated the underlying discrimination claim.
6. Publish limitations with every headline number.
Any public report based on contributor data should state the number of responses, collection period, missing-data treatment, self-selection limitation, and relevant geographic or issue concentration.
7. Never mix advocacy with fabricated certainty.
The project's thesis may be that the system is failing people. The evidence still has to be presented in a way that allows a skeptical reader to inspect the underlying facts.