Diagnostic intelligence
Do not prescribe the fix until the evidence supports the diagnosis.
Creator performance problems are often multicausal. WITNESS separates the observed symptom from candidate causes before deciding which explanation deserves support.
Diagnostic dimensions
Exposure
Did distribution opportunity change?
Audience response
Did people who received the work consume or engage differently?
Audience composition
Did discovery, returning, follower, search, recommendation, or campaign source mix change?
Content mix
Are the compared works, formats, durations, or topics actually comparable?
Timing
Is the pattern explained by lifecycle stage, seasonality, reporting lag, or an event window?
Evidence quality
Does Signal Integrity or provider context lower confidence in any source?
Diagnosis requires discrimination.
A useful investigation asks which observations should be different if one explanation is true instead of another. WITNESS uses those discriminating signals to challenge the leading explanation instead of merely adding more charts.
Some diagnoses stay provisional.
If the evidence supports a distribution shift but cannot identify the hidden platform mechanism, WITNESS should say exactly that. The supported Finding remains useful without inventing inaccessible details.