Retention Autopsy — forensic diagnostics for long-form YouTube

Your audience leaves at a sentence, not a minute.

We cross-reference your last 20 retention curves with your transcripts and return the exact moments viewers quit — timestamped, quoted, machine-verified. Not a dashboard. A forensic report, in 48 hours.

Claim a design-partner slot See how it reads a curve 3 of 3 free slots open · then $1,500
EXHIBIT A

One video, read properly

AUDIENCE WATCH RATIO × VIDEO PROGRESS 20 VIDEOS ANALYZED · THIS ONE: 14:22
CLIFF · 05:27–05:58−7.1% watch ratio in 31s
“…before we get into it, let me tell you about today’s sponsor. They reached out and honestly…” — the same read, at the same relative position, in 7 of your last 9 sponsored videos.

Hover or tap the markers. Illustrative curve — your report is built from your channel's real per-second data.

EXHIBIT B

What Studio shows you vs. what you get

YouTube Studio

  • One curve, one video at a time
  • "Viewers are leaving at 5:30" — no why
  • No transcript alignment
  • No memory across your catalog
  • You, squinting at 20 graphs on a Sunday

Retention Autopsy

  • Your last 20 videos, one analysis
  • The drop, the timestamp, and the sentence
  • Patterns that repeat across videos — and the videos that break them
  • Three severity lenses: vs. your video, vs. your channel, vs. similar videos on YouTube
  • A ranked, quoted, verified report + 45-min walkthrough

AI wrote you a channel audit? Cute. Ours has to survive an audit itself.

The evidence rule — Exhibit C
EXHIBIT C

The evidence rule

Detection is deterministic code, not vibes: drop events are extracted from your per-second curves mathematically, then cross-referenced with your transcripts. AI drafts the findings — and then a validator checks the AI's homework in code. A finding only ships if:

Findings that fail any check are deleted before the report renders. What reaches you is only what the data can defend.

EXHIBIT D

Anatomy of a finding

HIGH SEVERITY

The apology-intro is your most expensive habit

PATTERN  · intro_drop · 6 of 8 eligible videos MEDIAN   · −11.3% watch ratio inside the first 40s EXCEPTION · 2 videos that skip it hold +9.1% at the same mark
“hey guys, sorry this one took so long to come out, it's been a crazy few weeks…”

Action: cold-open the claim your title promised. Your two pattern-breaking videos already prove it works on your audience — this isn't generic advice, it's your own A/B test, found.

Every finding in your report carries this structure: the pattern, the receipts, the counter-examples, the move. Read a full sample report (synthetic demo data, clearly marked — real case studies replace it as design partners come aboard).

PROCEDURE

How it works

Connect — 15 min

A short call. You grant read-only analytics access live on screen; we pull your last 20 long-form videos. No passwords, no channel permissions.

Analysis — 48 h

Curves are resampled, drop events extracted, patterns mined across videos, findings drafted and machine-verified. A human (me) QAs every claim before it ships.

Walkthrough — 45 min

You get the report, then we read it together: what to fix first, what to leave alone, and which of your own videos already prove the fix.

Access & data, plainly

  • Read-only. Analytics and video metadata scopes only. We cannot edit, upload, or touch your channel.
  • Minimal-permissions mode. Transcripts can come from your public captions — no caption-API scope at all, if you prefer.
  • 30-day self-destruct. Your raw data is purged automatically after 30 days (or on request, immediately). We keep the findings, never the curves.
  • 7-day tokens. Access expires on its own within a week — by Google's design, and fine by us.
FIT

Built for some channels. Honestly not for others.

Strong fit

  • Long-form (8+ min), scripted or semi-scripted
  • 100k–2M subs with a business behind the channel
  • Education, finance, tech, video essays, documentary
  • Sponsor reads you'd like to stop paying for twice

Not for

  • Shorts-first channels — no long-form curves to read
  • Music & compilation channels — retention is structural, not scripted
  • Channels under ~20 published long-form videos
  • Anyone wanting a growth-hack dashboard — this is a diagnosis, done by hand
ENGAGEMENT

Pricing

3 SLOTS · THEN GONE Design partner $0

The full Autopsy, free, in exchange for a short testimonial and permission to anonymize findings. First three channels only — this offer does not come back.

Claim a slot
Single Autopsy $1,500

Your last 20 long-form videos: ranked findings, verified evidence, per-video annotated charts, and the 45-minute walkthrough.

Book
Quarterly $3,600/yr

An Autopsy every quarter, plus deltas against the last one — did the fix actually move the curve? The only tier that answers that.

Book

Run an agency? The pipeline white-labels: your brand, our evidence layer, $600/report at volume. First one free on a current client.

OPERATOR

Who's behind this

I'm Zane. I trade quantitatively — reading time series for signal, under real stakes, is my day job. The Retention Autopsy applies the same discipline to the one time series every creator owns and almost nobody reads properly: the retention curve. Detection is code, claims are verified, and I personally QA every report before it reaches you.

It's deliberately a service, not a SaaS: five channels at a time, done carefully.

— zaneyu2005@gmail.com

Twenty curves. Every sentence that costs you viewers.

Three design-partner slots, then $1,500. If your channel fits, the first report is the cheapest it will ever be.

Claim a design-partner slot Read the sample report