Veritasium

• Massive established audience and reach: 21.0M subscribers and 4.40B total views provide strong algorithmic momentum and off-platform authority.
• High average views per upload (6.69M) indicates durable topic selection and strong long-tail performance across 520 videos.
• Clear niche positioning (science/education) with broad topical range (physics, biology, chemistry, astronomy, math) supports both evergreen search and suggested traffic.
• Top videos demonstrate strong curiosity hooks and stakes-driven framing (e.g., “cut the green rope?”, “faster than light,” “weeks away from disaster”) that reliably generate 10M–39M view outcomes.
• Consistency and cadence are excellent (avg 4.5 days between uploads; score consistency 90, frequency 100), which sustains audience habit and recommendation velocity.

• Engagement score is relatively low (50) despite large absolute likes/comments; several top videos show modest engagement rates (e.g., 1.4–2.13), suggesting room to increase viewer interaction per view via stronger prompts/structures.
• SEO score is middling (50); keywords are broad and may not be tightly mapped to specific recurring series, searchable problem statements, or high-intent queries.
• Packaging variability: some top-performing titles are extremely strong hooks, but others can be improved with clearer benefit/stakes language to reduce “scroll past” risk at scale.
• Broad topic spread can dilute returning-viewer expectations if not organized into clear, repeatable series pillars (e.g., physics paradoxes vs. engineering vs. internet/tech failures).

• Run systematic title/thumbnail experiments on the next 10 uploads: keep topic constant but test 2–3 packaging variants to lift CTR; even a small CTR gain on a 6.69M average-view baseline produces outsized incremental views.
• Create 3–4 recurring series pillars and label them consistently in titles/thumbnails (e.g., “Misconceptions,” “Physics Paradoxes,” “Engineering Disasters,” “Impossible Experiments”) to improve viewer expectation and suggested-video chaining.
• Add stronger first-30-seconds structure: faster problem statement + visual demonstration upfront, then explanation; use the top-video pattern (“What happens if…”) as the opening blueprint to improve retention.
• Build sequel chains from proven winners: produce follow-ups to the highest-view topics (rope/forces, faster-than-light framing, critical infrastructure failures, antimatter scale) with explicit connective titling (Part 2 / “Again, but…” / “The next question is…”).
• Improve evergreen search capture by rewriting descriptions with 3–5 specific searchable questions per video and adding chapter titles that match common queries (e.g., “why does…”, “how can…”, “what if…”).

• Add a tighter on-video subscription promise tied to the viewer’s current interest (e.g., “If you like counterintuitive physics experiments, subscribe—new paradox every week”), rather than a generic CTA; aim to convert the massive view base into more subs.
• Use end screens to push into a ‘starter path’ playlist (3–5 videos) instead of a single recommended video to increase session time and subscriber conversion.
• Pin a top comment that invites a specific action (vote on next experiment/topic, submit misconceptions) to raise community participation and repeat viewing; current average comments (8.4K) can be leveraged for community-building.
• Package periodic ‘best-of’ or ‘essential concepts’ playlists for new viewers (e.g., “Start Here: 10 Mind-Bending Physics Ideas”) to reduce overwhelm from a 520-video library and improve new-subscriber onboarding.
• Increase community posts tied to upcoming uploads (teasers, polls) to prime returning viewers and improve early velocity, which helps recommendations.

• SEO uplift: with SEO scored at 50, better query alignment (titles, chapters, playlists) can add a meaningful evergreen traffic layer on top of already-strong suggested traffic.
• Engagement lift: top-video engagement rates (1.4–3.32) suggest headroom; structured prompts (prediction moments, viewer experiments, comment challenges) can raise interactions per view.
• More tech/infrastructure storylines: “The Internet Was Weeks Away From Disaster…” (14.7M views, 2.83 engagement) shows strong appetite for high-stakes systems narratives beyond pure physics.
• Health/industry risk stories perform well (“asbestos problem…” at 10.1M views, 3.32 engagement); expanding this pillar carefully can diversify audience and attract premium sponsors.
• Library optimization: 520 videos means significant compounding gains from updating end screens, cards, playlists, and descriptions across the back catalog.

• Topic dilution risk: covering “anything else we find interesting” can reduce repeat-view predictability if not anchored to a few recognizable pillars.
• Packaging saturation: at this scale, small declines in CTR/retention can cause large absolute view losses; maintaining high-concept hooks without becoming clickbait is a balancing act.
• Production strain: maintaining a 4.5-day cadence for high-quality science content can pressure quality control, which could impact retention and trust if rushed.
• Overreliance on browse/suggested: if SEO remains moderate, algorithm shifts could impact discovery; diversifying into stronger search and playlist flows reduces this risk.

You have the rare combination of scale (21M subscribers), consistency (4.5-day average cadence), and proven repeatable hit formats (multiple 10M–39M view videos). The biggest upside now is operational: tighten series pillars, systematize packaging tests, and improve SEO/engagement mechanics so each upload converts a slightly larger share of impressions and viewers—small percentage improvements at your baseline translate into massive growth.