Algorithmic Positioning Audit — The Dossier (@thedossier_in)

Sample note

Audit covers 60 of 80 known videos (75%); the other 20 could not be fetched (bot challenge). The sample contains 24 long-form videos and 36 shorts, spanning upload dates 2025-04-02 to 2026-05-05 — the channel’s full operating window. Because the missing 25% is randomly blocked rather than chronologically biased, the cadence and ratio numbers below are directionally correct but absolute counts undercount uploads by roughly one-third. All claims below are derived only from channel.json, the 60 info.json files, and summary.jsonl; no transcript, comment, or thumbnail data was consulted.

Cadence audit

Trailing 6-month uploads: 28 videos (~4.7 per month). Trailing 12-month: 56 videos (~4.7 per month). The two windows match, meaning there has been no sustained acceleration over the year — only a shift in composition (see Shorts audit).
Trailing 90 days: 25 uploads (7 long-form + 18 shorts). Prior 90 days: 3 uploads (all long-form, 0 shorts). Total volume is up 8x, but almost all the new volume is shorts.
Longest gap in last 12 months: 49 days between long-form uploads. Median long-form interval: 14 days. Mean long-form interval: 16.5 days. The long-form pipeline runs at roughly bi-weekly cadence with frequent month-long droughts.
Monthly long-form output is lumpy: the channel posts 1 long-form in 7 of 14 months in the sample, and concentrates effort in bursts (e.g., 4 long-form in 2026-04 [yLEakbf5gpI] [nFIGdsXAYpA] [Q9wui2puJdE] [3uk0G_LNd7A] after only 1 in 2026-02 and 1 in 2026-03).
Day-of-week (long-form, 24 videos): Sat 9, Sun 5, Mon 5, Wed 3, Thu 1, Fri 1, Tue 0. Long-form is a weekend-anchored channel — 58% of long-form ships Sat/Sun/Mon. Shorts DOW (36 videos): essentially flat (Mon/Tue/Wed/Fri/Sat all at 6). The long-form weekend pattern is real; the shorts pattern is “whenever.”
Two notable gaps in 2025-11 to 2026-02: 4 months produced only 6 uploads total (all long-form, no shorts), bracketed by a heavy shorts burst in 2025-09 to 2025-10 (17 shorts) and a heavier one in 2026-03 (13 shorts). Output is binge-and-rest, not steady.

Length strategy audit

Long-form duration distribution (38 videos with duration >=90s in info.json; quartile and decile analysis uses the 24 in summary.jsonl which carry view_z_score):

Bucket	Count
0-5 min	15
5-10 min	2
10-15 min	1
15-20 min	4
20-25 min	3
25-30 min	0
30-40 min	1
40-50 min	3
50-60 min	3
60+ min	6

Bimodal distribution. Two clusters: short-cut packages 0-5 min (15 videos — including mis-categorized “long-form-but-tiny” pieces like [nFIGdsXAYpA] 293s, [W4I42zE8qTA] 328s, [Gs38SCGxPGA] 484s) and a heavy long-form tail at 20+ min. The 5-15 minute “discoverability sweet spot” range holds only 3 videos. The channel either makes a deep dive or a clip; the middle is empty.
Top-vs-bottom decile by view_z_score (24-video set):
- Top decile median length: 2,992s (~50 min) — [yCR38NmeCpY] 46 min z=+2.44, [oYTJaHw4na8] 50 min z=+1.74.
- Bottom decile median length: 1,015s (~17 min) — [6Ix6Tt4I47A] 12 min z=-1.81, [atLAW_yl70Q] 17 min z=-1.74.
Top quartile vs bottom quartile confirms the pattern: top-Q mean 3,028s (~50 min), bottom-Q mean 865s (~14 min). The longest videos on this channel outperform the medium-length videos by a wide margin. The channel’s algorithmic edge is watch-time density on long videos; short long-form is the weakest format.
The single longest video [6TVBySERPzE] (110 min) and the longest “shorts” [NHTNJPcvO0Q] 173s, [Y6pF1fIZu_c] 161s sit just above and just below the 3-minute Shorts cap — three of the 36 “shorts” exceed 144s, near the YouTube Shorts ceiling.

Shorts audit

Volume: 36 of 60 sampled videos are shorts (60%). Mean short duration: 79s. Range: 21s-173s.
Trend: Shorts went from 0 in 2025-04 to 2025-07, to a 9-short burst in 2025-09, an 8-short burst in 2025-10, a 4-month dormancy (Nov 2025-Feb 2026 with zero shorts), then 13 shorts in 2026-03 and 4 more in 2026-04. Last-90-day shorts: 18. Prior-90-day shorts: 0. Shorts strategy is binge-mode, not always-on.
Shorts to long-form funnel: nonexistent. Every short in the sample has an empty description (36/36, mean 0 chars). Zero shorts contain a YouTube URL in the description. Only 1 of 36 shorts has any CTA at all in the script extract ([JZAr30UgX_c] says “tell us”). Shorts on this channel terminate the funnel; they do not redirect attention to long-form. This is the single largest structural leak in the system.
Title hygiene is fine for shorts: 30 of 36 carry #shorts in the title; the 6 that don’t are mostly Hindi titles where the hashtag was omitted.
Three “shorts” ([NHTNJPcvO0Q], [Y6pF1fIZu_c], [Flc5mIdH6Ug]) exceed 2:30, brushing the Shorts shelf cutoff — these may be silently re-classified by YouTube as long-form, splitting their reach.

Playlist audit

Custom playlists: zero. channel.json reports playlist_count: 2, and both entries are the auto-generated The Dossier - Videos (32) and The Dossier - Shorts (48). No info.json carries a playlist or playlist_id field (0/60).
Coverage: 0% of long-form videos are organized into a topical playlist. The channel’s actual topic clusters from summary.jsonl (politics_govt, geopolitics, religion_society) would each support a 5-10 video playlist today — e.g., the Operation Sindoor / Pakistan-conflict cluster ([WV3dCPpNOrk], [Gs38SCGxPGA], [DNHWfGBfUiM]), the Waqf / Islamic-law cluster ([fjGkaPLP_KA], [1E6So2gYgP8]), the judiciary cluster ([zv2HCSdsiJw], [Hdrj7Q21Fm0]), and the Ex-Muslim interview cluster ([oYTJaHw4na8], [t2JcdxBweSw], [nFIGdsXAYpA]).
Consequence: no “Next Up” surface to feed session watch-time, no playlist landing pages for SEO, no end-of-video continuation that YouTube can pre-load.

End-screen and card audit

info.json does not directly expose end-screens or cards (yt-dlp surfaces neither field). Chapters and description links are used here as proxies for the kind of structural enrichment the channel is doing on its own videos.
Chapters on long-form: 13 of 24 (54%), of which 11 are real named chapters and 2 are placeholder/auto chapters. 11 long-form videos ship with zero chapters — including high-stakes recent uploads like [3uk0G_LNd7A] (18 min), [Ipd9sXG8dLI] (19 min), [1E6So2gYgP8] (23 min), [WV3dCPpNOrk] (41 min), [yCR38NmeCpY] (46 min), and [nC4cyUEVDek] (51 min). All of these are long enough that chapters would meaningfully aid retention and replay.
Description-level cross-links: 0 of 24 long-form descriptions contain a YouTube link to another video or playlist. There is no in-description funnel back into the catalog at all.
All 60 videos are playable_in_embed: True — no embed-blocking is happening, which is correct.

Premiere usage

live_status is not_live on all 60 videos. No was-live broadcasts in the sample.
release_timestamp is set on 6 videos ([Ipd9sXG8dLI], [t2JcdxBweSw], [nFIGdsXAYpA], [1E6So2gYgP8], [3uk0G_LNd7A], [Q9wui2puJdE]), all from 2026-01 onward. These six are the only candidates for scheduled premieres in the sample — all long-form, all from the most recent 4 months. The channel has begun experimenting with scheduled releases but it is not yet a consistent practice (6 of the 24 long-form, 25%).

Seven algorithmic changes ranked by leverage

Add a closing card and pinned-comment link from every short to one specific long-form video. Currently 36/36 shorts have empty descriptions and 35/36 have no CTA, so a single short never sends a viewer to a long-form upload. Expected effect: convert a fraction of the channel’s largest viewer surface (shorts) into long-form watch-time, which is the format where view_z is highest. Effort: low.
Build at least four topical playlists from the existing catalog (Operation Sindoor / Waqf / Judiciary / Ex-Muslim) and set each long-form video to auto-play the next in the playlist. The channel has zero custom playlists today and zero long-form descriptions link sideways. Expected effect: surfaces a session-extension path YouTube can chain, raising average view duration per session without new production. Effort: low.
Ship chapters on the 11 long-form videos that lack them, prioritizing the 20+ minute videos [WV3dCPpNOrk], [yCR38NmeCpY], [nC4cyUEVDek]. Chapter markers exist on 54% of long-form today; the missing half includes most of the longest videos. Expected effect: improves seek/replay and exposes a key-moments strip in search, addressing the format (long-form) where the channel’s top decile already wins. Effort: low.
Lock a long-form cadence of one upload every Saturday. Long-form is already weekend-skewed (9 of 24 on Sat) but the median interval is 14 days with a 49-day gap; the algorithm rewards predictability. Expected effect: closes the 30-49 day droughts that currently kill the algorithm’s ability to model the channel’s frequency. Effort: medium (requires either a buffer of finished videos or a smaller-but-steadier format).
Convert the next 12 long-form uploads into scheduled Premieres rather than direct publishes. Only 6 of 24 long-form so far appear to have used release_timestamp. Expected effect: creates a notification surface and a synchronous-viewing spike, which feeds early-watch-time signal — the metric YouTube uses to decide initial recommendation push. Effort: low.
Trim the three over-length “shorts” ([NHTNJPcvO0Q] 173s, [Y6pF1fIZu_c] 161s, [Flc5mIdH6Ug] 155s) below 150s on re-uploads, and cap all future shorts at 150s. Their current length sits at the Shorts/Long-form border where YouTube’s classifier behavior is ambiguous. Expected effect: ensures the shorts surface keeps them, removing format-class risk. Effort: low.
Reduce the 5-15 minute long-form gap by promoting the 0-5 min “mini long-form” pieces ([nFIGdsXAYpA] 293s, [W4I42zE8qTA] 328s, [Gs38SCGxPGA] 484s) into either >15-min explainers or true shorts. The distribution is bimodal with a hole in the 5-15 min range, and the bottom-decile videos by view_z all sit at the lower edge of long-form (median 1,015s). Expected effect: stops shipping into the format-length band that empirically underperforms on this channel. Effort: medium.