# Evidence decisions working packet

Seven new reader jobs and one deepened existing interview guide. Each bilingual guide has complete contextual cases and a private blank record. All money, samples, p-values, notes, people and outcomes in the exercises are constructed. No real accounts, tax records, contracts, sponsorships, interviews, preferences or payments are represented.

The on-page cash/time/contribution calculator is a local planning aid. Fixed cash excludes the entered variable unit costs. An unpaid-time value is an assumption, rounded once half up to cents. Empty is unknown; a declaration against double-counting is unverified. A positive unit contribution is required for its threshold calculation, and capacity/demand remain separate. The calculator supplies no actual price, tax treatment, accounting profit or demand forecast.

Run node check-examples.mjs from this extracted folder using an existing Node installation. It uses only included JSON and Node built-ins. It checks constructed integer-cent boundaries, cost allocations, declared sample-precision arithmetic, a finite-family error exercise and note IDs. It does not run a real statistical test or validate data, model assumptions, sampling, permission, legal disclosure, human interpretation or business value. The app's input-validation tests are separate from this arithmetic checker.

US FTC guidance and the German May 2025 Medienanstalten guide/current statute have distinct scope and wording. No universal hashtag or legal clearance is offered. The NIST sample-size examples use a rounded normal value near 2; the original precision exercise explicitly uses 1.96. Nominal precision is not representativeness, power, achieved confidence or a replacement for an actual study design. NIST's multiple-comparison example is ANOVA-specific; the general inequality is used only with its stated valid-test conditions. ASA's 2016 statement summary is historical primary guidance, not a newly published benchmark.

The four interview notes deepen the existing source-to-story task. They do not add a topic, participant, independent reviewer or research study. Keep contradictory and unobserved notes visible and preserve actual source/consent rules for any real work. No actual recruitment, outreach, publication, message, service purchase or new automation occurred. The packet supplies no observed reader preference or customer-outcome evidence.

The identity manifest excludes itself and the ZIP to avoid circular hashes. The archive includes all source files and that manifest. Keep completed records private and minimal; the packet is not a permission or consent form.