THE DOSSIER · GROWTH ANALYSIS
SNAPSHOT · 2026-05-17
Channel growth dossier · India · politics, society, religion, law

The Dossier: 18,600 → 100,000 in 90 days

An end-to-end analysis of the channel @thedossier_in. One snapshot. Eight subagent reports. A 90-day experiment program with kill criteria, sequenced into a 12-week plan.

Subscribers
18,600
target: 100,000 · 5.4×
Videos sampled
60
24 long · 36 shorts
Cumulative views
838,902
across the catalogue
Median views
11,336.5
long-form · shorts 1,446.5
The gap to 100k
Where the channel sits today
TODAY 18,600 TARGET · 90 DAYS 100,000 +81,400 required
Subscribers held Gap to 100k
Source: yt-dlp channel_follower_count
The three leaks that cap growth
Where the channel is bleeding subscribers
Three binding constraints
The leaks that cap growth regardless of content quality
CONSTRAINT 01
Shorts funnel sealed shut
36 / 36
shorts with empty descriptions
0 link back to long-form. The highest-velocity surface terminates the funnel.
CONSTRAINT 02
Binge-and-rest cadence
49d
longest long-form gap (last 12mo)
The recommender can't model frequency. Audience asks 'where have you been?'
CONSTRAINT 03
Length distribution skips the winners
0
videos in the 25–30m winner band
5/6 top long-form sit at 25–50m. The channel ships 0 there.
Synthesized from algorithmic-positioning, content-forensics, audience-community
The sealed shorts funnel
Where the channel's highest-velocity surface dead-ends
Shorts feed 36 videos · top z=2.78 Long-form where subs convert BLOCKED 0 / 36 link CURRENT STATE • 36 / 36 shorts: empty description • 0 / 36 shorts: link to a long-form video • 1 / 36 shorts: any scripted CTA in audio PROPOSED FIX (#1) • 3-line description template • 2-sec outro voice tag • Pinned long-form link
algorithmic-positioning report
Long-form length distribution
A hole where the winners live
0 3 6 1 0–5m 2 5–10m 1 10–15m 4 15–20m 3 20–25m 25–30m 2 30–45m 5 45–60m 6 60m+ LENGTH BUCKET N VIDEOS
Winner band (25–45m) Graveyard (5–15m) Other
Source: data/videos/summary.jsonl · 24 long-form videos
The audience bought a person, not a publication
Mentions in a 702-comment filtered sample
"Kaushlesh" 106 mentions of the host's name "Dossier" 11 mentions of the channel's name ~9.6× MORE MENTIONS OF THE HOST
audience-community + brand-identity reports

What changes in the 90-day program

Today By day 90
Cadence Binge-and-rest; 49-day long-form gaps; shorts in monthly bursts. Daily short (Mon–Fri), weekly numbered recap (Wed), Saturday-7pm long-form premiere.
Shorts funnel 36/36 shorts have empty descriptions; 0 link to long-form. 3-line description template + 2-sec voice tag on every short pointing to a parent long-form.
Long-form length Bimodal — 15 of 24 sit at 0–5 min; 0 in the proven 25–30m winner band. Editorial rule: every long-form is ≥25 min or recut to a short. 5–15 min band killed.
Identity Marquee says 'The Dossier'; audience addresses the host as Kaushlesh (~9.6× more often). Channel renamed 'Kaushlesh Rai | The Dossier'; locked cold-open and sign-off in every video.
Packaging Host-anchor split-portrait recipe used inconsistently; bottoms break it. One thumbnail spec, no question marks, host-right + antagonist-left + bold red/yellow overlay.
Conversion No pinned CTAs; no email/sources artefact; donation buried in description. Pinned 4-step action ladder + 1-page sources PDF on every investigation episode.
Per-cohort performance distribution
Winners and losers, separated by format
Long-form: per-cohort view z-score
n = 24 videos · quartile cut at ±6
-2 -1 0 +1 +2 peak +2.44 floor -1.81
Top quartile Bottom quartile Middle
Source: data/videos/summary.jsonl
Shorts: per-cohort view z-score
n = 36 videos · quartile cut at ±9
-2 -1 0 +1 +2 peak +2.78 floor -1.83
Top quartile Bottom quartile Middle
Source: data/videos/summary.jsonl

Start where the leverage is

If you read three things, read these in order. Everything else compounds on top of them.

  1. Synthesis — the 90-day roadmap, three binding constraints, the 12 ranked experiments, and the math on what 100k actually requires.
  2. Growth Experiments — each of the 12 with hypothesis, intervention, success metric, kill criteria, dependencies.
  3. All reports — the seven analysis subagents with the underlying evidence.
Source data acquired via yt-dlp. Channel under analysis: @thedossier_in · Snapshot date: 2026-05-17 · 60 of 80 videos sampled (YouTube bot challenge blocked the remainder). Every metric traces back to a file in data/.