Evidence Base

Contradictions register and confidence ratings across all findings. Confidence reflects source quality, not importance — a "weak" finding may still be the best available answer to an important question.

Contradictions Register

Where benchmark sources disagree, both sides are preserved. These are real measurement differences — each study uses different datasets, time windows, and metric definitions.

Evidence Confidence Index

Strong

Multiple independent sources with disclosed data, pointing the same direction.

• Posting frequency sweet spot: 3-5/week^[1][3][2]
• Carousel engagement leadership^[2][1][6]
• Collab post engagement lift: 2.7x for 10K-50K brands^[7]
• UGC conversion impact: 60% of shoppers always look for customer visuals^[8]
• Instagram SEO: captions > hashtags for search^[5]
• First impressions form in ~50ms^[9]

Moderate

One good source with data, or multiple sources pointing the same direction without clean independent replication.

• View-to-follow rate ~0.01-0.09% — two data points only^[10][11]
• Broadcast channel join rate ~0.25-1% of followers — directional, from public profile analysis^[12]
• Series content improves follow conversion — no clean A/B, but Instagram launched Linked Reels specifically for this^[13]
• Adjacent-niche partners > same-niche for collab reach — inference from mechanics, not direct test^[14]

Weak

Anecdotal, vendor-reported, single case study, or inferred from mechanics rather than measured.

• Live commerce results (4x sales, 5x comments) — Sprii vendor case studies^[15]
• ManyChat conversion rates (85% registration) — vendor success stories^[16]
• Broadcast channel driving 20% of DTC sales — single Megababe case^[12]
• "Too narrow" threshold for niche accounts — pattern observation, no published threshold [uncited — from synthesis]
• UGC conversion lifts 100%+ — vendor case studies^[17]