Public experiment register
Test small.
Write down everything.
No secret sauce. Each record names the variable, measurement window, evidence, and what the result cannot prove.
RB-EXP-001Measuring
Direct titles vs. brand-led titles
Direct, descriptive titles may establish clearer query alignment than brand-led titles on a young publication.
Protocol and limits
Hold the surrounding page shell as stable as practical, record changes before measurement, and do not extend a site-specific observation into a universal ranking rule.
- Signal
- Impressions + query mix
- Window
- 28 days
RB-EXP-002Queued
Definition-first glossary pages
Putting the plain-language answer first may improve usefulness without forcing every entry into long-form article theater.
Protocol and limits
Hold the surrounding page shell as stable as practical, record changes before measurement, and do not extend a site-specific observation into a universal ranking rule.
- Signal
- Indexation + engagement
- Window
- 42 days
RB-EXP-003Queued
Visible evidence labels
Readers may navigate more confidently when source depth and claim limits are visible before the body copy.
Protocol and limits
Hold the surrounding page shell as stable as practical, record changes before measurement, and do not extend a site-specific observation into a universal ranking rule.
- Signal
- Navigation paths
- Window
- 21 days
Lab rules
01Reversible
Every intervention should be easy to undo or isolate.
02Observable
Define the signal before looking for a result.
03Bounded
Write down what the experiment cannot prove.
04Recorded
No clean-up that erases an inconvenient outcome.