Caveats for the interactive-tool-mechanisms brief: lead on mechanism evidence (peer-reviewed, independent); treat vendor outcome stats (52.6% / 88% / 47.3%) as marketing

The honest summary of source quality across this brief:

• The strongest independent evidence sits at the mechanism level — peer-reviewed cognitive and behavioural science.
• The business-outcome statistics ("interactive content gets 2× engagement," "52.6% more engagement," "4-5× more pageviews") almost all trace back to vendors that sell interactive-content platforms — ion interactive (now Rock Content), Demand Metric (vendor-sponsored), Mediafly, Outgrow, Ceros, SnapApp.
• The single most-quoted source (Demand Metric 2014 Content & the Buyer's Journey Benchmark Study — vendor-sponsored online opinion survey of 185 marketers; the "2× engagement" headline rounds 70%/36%) is explicitly sponsored by ion interactive — n=185 marketers, opinion survey, not behavioural research.
• The Mediafly 52.6% / 88% figures (Mediafly / Demand Metric: "Interactive content shows 52.6% higher engagement than static; buyers spend 13 vs 8.5 minutes" — vendor sources, treat as marketing not fact) are vendor-sourced.
• The Pham 2024 brand-innovativeness → expertise → WOM finding (Pham et al. (2024), Australasian Marketing Journal — brand innovativeness has positive indirect effect on positive WOM, mediated by perceived brand expertise) is single-source peer-reviewed; the calculator-specific application is theory-grounded, not directly measured.
• The named-tool backlink magnitudes (Named-tool backlink magnitudes (Ahrefs-indexed, directional): CoSchedule Headline Analyzer 16K+/3.6K+; Coolors 154K+/5K+; ABV calc 2.3K/190+; Adobe Shortcut Mapper 280+/130+; Ahrefs free backlink checker 1M+) are Ahrefs-indexed and directional, not audited.
• The AI-engine forward-looking line (Forward-looking: AI answer engines may increasingly surface and cite useful tools, with brand mentions/links correlated with AI-overview visibility — early and contested) is explicitly Directional-Speculative.
• The Zeigarnik memory claim is largely non-replicated (CORRECTION: 2025 meta-analysis (Nature, Humanities & Social Sciences Communications; 38 Zeigarnik studies) — NO overall memory advantage for unfinished tasks (Cohen's dz ≈ 0.15)) — use the goal-gradient / Ovsiankina cousins instead.

Practical posture for Candid: lead on mechanism evidence; treat outcome stats as marketing. See R1 — When recommending an interactive tool, LEAD on peer-reviewed mechanism evidence (goal-gradient, self-reference, IKEA, reciprocity, anchoring) — NOT vendor "2× / 47% / 16.9×" stats.

Cross-brief: This is the mechanism-evidence sibling of Caveats for the customer-facing-calculators brief: every conversion-lift figure is unproven; nearly all are vendor-self-reported. Together the two briefs form Candid's complete position: the calculator brief takes down the vendor stats; this brief supplies the independent positive case.