# Channel engagement comparison recipe

A source-level analysis contract that separates comparable activity and rate measures from raw audience-sensitive counters.

## Calculation

- Mode: source_comparison
- Formula: Per source: posts/window days; forwarded posts/all posts; medians of captured metrics and per-view rates; top domains by distinct logical posts.

## Query

Compare [selected channels] from [start] to [end]. Report posts/day, forward count and ratio, median captured views/forwards/reactions with sample size, median reactions/view and forwards/view, and top three domains. Name zero-post sources, cite representative posts, and do not rank channels by raw medians alone.

## Inputs

- **Sources:** The exact channels whose behavior will be profiled.
- **Window:** One shared publication interval and timezone.
- **Comparison fields:** Activity, forwarding, captured medians, relative rates, and domains.
- **Decision context:** The research question that determines which differences matter.

## Expected output

- One statistical row per source
- Common window and post counts
- Raw medians and measured sample size
- Relative per-view rates
- Domains, citations, and missing sources

## Coverage checks

- [ ] Use one date window across sources
- [ ] Require sufficient samples for medians
- [ ] Do not rank raw counters across different audiences
- [ ] Do not call unmarked posts original
- [ ] Read cited content before judging source quality