Video summary

YouTube was testing my videos wrong (so I fixed it)

Main summary

Key takeaways

Business

Executive summary (business-focused)

  • The creator audited their YouTube channel and found that YouTube’s recommendation system was showing videos to the wrong audiences, which led to low click-through rates (CTR) and poor “suggested videos” performance.
  • They use YouTube Studio analytics reports to diagnose the mismatch, then apply a 3-step “title-as-signals” playbook to give YouTube clearer signals about who the video is for and what the video is about—specifically to improve Suggested videos CTR.

Framework / playbook: “Give the algorithm better clues” (3-step title fix)

Goal: Improve Suggested CTR by making video metadata (early title text + first description line) unambiguous.

Fix #1: Cut vague words; name the actual topic

  • Avoid: “tips, tricks, secrets, hacks, growth”
  • Replace with specific “what”
    • Example: “skin-care tips and tricks” → “double cleanse routine for oily skin”

Fix #2: Make the title about the intended viewer (“who”)

  • Don’t just say “my/me/i” (that’s not the real issue); ensure there’s clear audience identity.
  • Two acceptable ways to signal “who”:
    1. Name it outright
      • Example: “small channels fail” (direct audience)
    2. Use a situational label
      • Example: “My first 30 days on YouTube” (implicitly signals “new creators”)
  • Avoid titles that state neither clear “who” nor “what”
    • Example: “my workout routine tells me” (unclear audience + unclear value)

Fix #3: Drop broad “killer words” / overly competitive categories

  • Avoid broad singles/genres that big channels dominate (e.g., “money,” “success,” “video,” “recipe”).
  • Narrow to a specific person + scenario
    • Example: “how to make money on YouTube” (competes with massive channels)
    • Replace with: “30-minute sheet pan dinners for busy parents” (much narrower targeting)

Diagnostic process (YouTube analytics reports used as “operations tooling”)

Step-by-step workflow

  • Pick a low-performing video.
  • Open YouTube Studio → Analytics → Reach.
  • Drill into:
    1. “Content suggesting this video” (who YouTube recommends it alongside)
    2. Source-level CTR breakdown (Browse vs Suggested)
    3. Impressions CTR over time/level (a line/graph indicating seeding difficulty)

What each report is meant to reveal

  • Report 1 (Next-to-which-videos): “Wrong room” detection

    • If top suggested placements include unrelated niches, CTR collapses because the wrong audience is being seeded.
    • Heuristic: if 3 of the top 5 suggested videos are off-niche → “YouTube has not received enough clues.”

  • Report 2 (Source CTR mismatch): “Fans love it, but new fans won’t bite”

    • Key idea: CTR is not one number—it’s different per traffic source.
    • Interpreting mismatch:
      • High Browse CTR = title/thumbnail resonates with existing viewers
      • Low Suggested CTR = algorithm struggles finding the right new audience

  • Report 3 (Impressions vs CTR pattern): “Algorithm uncertainty”

    • As impressions rise, average CTR often drops because YouTube tests more audiences.
    • Healthy pattern: steadier line
    • Unhealthy pattern: zigzags/large swings → YouTube “gets it right and wrong,” struggling to identify audience fit.

Concrete metrics & observed thresholds (examples from the video)

Observed “Suggested CTR” problem (example video)

  • “Clearly related to content creation” (but still low):
    • “0% click-through rate” for one suggested placement (shown as example outcome)
    • Another: 1.2% CTR with 31s average view duration
    • Another: 3 views from “unexpected content,” 31s average view duration
  • Takeaway: users shown via suggested placements were not the intended audience.

Observed source mismatch (example video)

  • Browse CTR: 8.1% (strong for known audience)
  • Suggested CTR: 1.9% (weak → algorithm isn’t finding the right new viewers)

Impressions→CTR variability examples

  • Mentioned CTR values in zigzag chart: 6.8%, 2.2%, 1.8%, 5.1%
  • Interpretation: large fluctuations indicate the recommender is testing multiple audience segments unsuccessfully.

Targets / timelines / operational guidance

  • Metadata changes take ~48 hours (up to ~48–72 hours) to process/propagate through YouTube systems.
    • Recommendation: “don’t go crazy”—be patient before judging results.

Actionable recommendations (what to do next)

  • Audit specific underperforming videos by checking:
    1. What videos YouTube suggests yours next to
    2. Browse CTR vs Suggested CTR
    3. Impressions CTR line behavior (zigzags vs steady)
  • Rewrite titles using a “who + what” structure, focusing on:
    • First ~40 title characters
    • First line of the description
  • Repackage old flops: changing title/metadata can revive videos (the creator notes old videos on an “old channel” restarted growth after years).

Investing/markets note (high level only)

  • No meaningful market/investing strategy is presented; focus remains on execution and recommendation-system optimization (metadata + audience targeting).

Presenters / sources

  • Presenter: The YouTube creator speaking in the video (name not provided in the subtitles)
  • Source system/tool: YouTube Studio (AI app studio) and YouTube’s AI recommendations

Original video