archive / annex / internal record
Telemetry Annex
Contact was made. Attention was not always retained. The numbers are accurate. They do not improve on reflection.
You found this. The record continues regardless.
Summary record / ~10 months
23
Works published
Posted into open air. The system registered each one and moved on.
38%
Contact above midpoint
9 of 23 works held past the halfway mark. The others did not collapse gracefully.
847
Peak views / single work
Autopsy #006. 23% contact. The most seen work is also the most documented failure.
0.7%
Response rate (average)
Presence without contact. The work was seen. It was not always encountered.
+22pts
EXP-007 current gap
Disruption variant holding. The only number in this table moving in the right direction.
4
Hours until dead
94% of views arrive in the first four hours. After that, the system has already decided.
Context
These numbers represent roughly ten months of work, one person, no team, no budget.
That context is accurate and also irrelevant to the shape of the curves.
The system does not make adjustments for effort.
It distributes based on signals it can measure.
The signals were frequently weak. The system responded accordingly.
Contact held — all works / ordered by midpoint retention
ordered by midpoint retention / all works with available data / floor = 18%
Failure log — works where contact was not made
| Ref | Midpoint | Views | Cause | Note |
|---|---|---|---|---|
| AUT-006 | 23% | 847 | Structure — setup gap | Most-viewed failure. The reach was fine. The structure failed it. |
| AUT-003 | 30% | 423 | Context — dead window | Right form, wrong moment. Filed under variables that are now controlled. |
| EXP-001 | 18% | 291 | No structure / baseline | The floor. Preserved to make everything else legible. |
| EXP-005-C | 29% | 502 | Timing — dead window | Contact was made with the distribution system, not an audience. The distinction matters. |
| EXP-002-daily | 31% | avg 380 | Volume — individual dilution | Daily output averaged 28% below the lower-volume condition per work. Total was comparable. Individual was worse. |
| AUT-005 ×5 | avg 26% | avg 441 | Structure — opening sequence | Five of six hook variants failed in the same direction. That's a pattern, not bad luck. |
Observation
The failure log contains more entries than the success record.
This is the correct ratio. Success is the rarer event.
A system that reports only the successes is not a record — it's a highlights reel
with the footage removed.
Distribution record / by platform
Primary platform
- 23 works distributed / ~10 months
- Contact above midpoint: 9 of 23 (39%)
- One effective contact window identified — others inconclusive or dead
- The system registered each work. Most were processed without further distribution.
- Reach collapsed without ceremony on the majority of outputs. No notification.
Secondary platforms
- Cross-posting: 8 works / inconsistent overlap period
- Average contact: 22% below primary — not a useful distribution channel at this scale
- Two exceptions where secondary outperformed primary. Neither was repeatable.
- The system registered the attempt and moved on. As did the audience.
On the contact window question
EXP-005 confirmed that one window is dominant for this account at this scale.
It was not the window that felt most convenient.
The inconvenience is now managed. The window is now controlled.
This is the most mechanical finding in the record.
It is also the finding that changed nothing about whether the work was any good.
Which is noted here because it is easy to lose.
Patterns confirmed / evidence-backed
What the data says
- The first seven seconds determine the shape of the contact curve in the majority of works reviewed.
- Contact lost before the midpoint almost always indicates a structural problem, not a content one. The audience rarely reached the content.
- Timing affects reach. Reach does not affect contact quality. These are different variables.
- The gap between best and worst variant in any controlled test is larger than anticipated. Every time.
- Presence without depth (audio mismatch case, AUT-002) is a distinct failure mode. Harder to read than a clean drop. No metric directly surfaces it.
- Success autopsies are harder to write. The evidence is less legible. The temptation to attribute causality is stronger when the outcome was good.
What did not help
- Volume without structural attention. More works at lower quality did not aggregate into equivalent contact. EXP-002 is clear on this.
- Posting into dead windows. The work was not the variable. EXP-005 confirmed. Now controlled.
- Assuming that what held once would hold again without analysis. AUT-004's success remained unexplained for three months. Still partially unexplained.
- Treating the absence of distribution as audience rejection. The system often simply did not distribute. These are not the same thing.
What held
- Keeping the record. The archive is not the failure. The archive is the only part of this that produces compounding value.
- The contact that did occur was meaningful. Small numbers. High signal per unit. Not scalable. Still true.
- EXP-007 is holding. Disruption variant gap at +22 points, day 8. This may close. It may not.
- The floor is now known. EXP-001 established it. Everything since has been measured against it. That measurement is useful regardless of the numbers.
Closing record
For the record
The numbers in this annex are not presented to perform honesty.
They are presented because the archive that selectively reports
is not an archive — it is a curated impression,
which is a different project with a different name.
This project's name acknowledges the error condition.
The annex is where the error conditions live in full.
You found it. The record continues.
annex / secondary record / access via topology / not the primary index