Methodology
What is measured, what is inferred, and what gets revised
The project is intentionally generous about showing its structure. Categories, experiments, and editorial recommendations all have different confidence profiles.
Experiments remain visible even when the outcome is embarrassing
The content model separates hypothesis, intervention, outcome, measurements, and confidence note so a failed test still improves the editorial system.
Daily misting failure
No durable improvement. Leaf surfaces looked better briefly after spraying, but edge crisping and decline pattern stayed functionally unchanged.
Humidity tray baseline
Slight reduction in edge crisping, but the bigger improvement came from more even watering and less afternoon heat exposure.
Mulch versus bare soil in a shade bed
The mulched zone held moisture longer and required fewer recovery waterings, with better frond density by the end of the test period.
Categorization rules
Plant profiles only store facts that can survive being reused in a card, table, filter, or related-content block. If a term appears repeatedly, it moves into the shared taxonomy layer rather than hiding inside prose.
Measured versus inferred
eFerns distinguishes four kinds of claims:
- Profile facts describe baseline plant requirements such as light range, water demand, and growth habit.
- Field observations describe repeated patterns noticed across maintenance cycles but not run as a formal experiment.
- Experiments describe a specific intervention with duration, measurements, and a confidence note.
- Recommendations are editorial judgments that must point back to either profile facts, field observations, or logged experiments.
Review cadence
Guides display both publish date and reviewed date. A reviewed date means the structured relations, taxonomy terms, and supporting examples were checked for consistency.
Editor safety rules
Summary and glossary fields are hard-limited in CloudCannon to preserve the grid. Top-level pages can use page sections. Reusable content types such as plants, guides, experiments, and glossary entries stay in their own collections so editors cannot accidentally turn structured data into layout fragments.