Aggregate observations from one organization · Historical trajectory · Not a probability forecast.

Research desk / 01

Read the depth.
Find your coordinates.

NADIR — The Black Box. Read one organization’s decline through eight measured axes.

8observed axes984monthly coordinatesOpen scale mode ↗
DECISION ATLAS

Every question, and what it came from.

Scale mode
Explore the same records in the question library.
One dot = one question
A line joins questions from the same scenario
Ctrl+wheel to zoom · Drag to rotate · Shift+drag to pan
THE WHOLE MAP

1,542

Questions on this map, gathered by the property of the organization each one reads. A line joins the branches of one scenario; every scenario so far sits inside a single property, so the lines make tight knots rather than crossings.

1,736 links across 496 questions · 1,046 have no line leaving them yet · distance on this map is not similarity or cause.

Nadir · The lowest point

This is how far it went

Approximately 18.58 million internal messages from one manufacturing and retail company, February 2016 through April 2026. This view traces the recorded trajectory toward the observed low, including narrowing communication and spreading silence.

Monthly messages 70.2×2016-08 436,944 messages → 2026-02 6,225 mentions
Monthly active speakers 12.1×2017-01 266 speakers → 2026-02 22 speakers
Peak channel inactivity 148 Highest month · of 3,466 channels
Median final-period ratio 0.562 Final 3 months ÷ preceding 3 months · 42.9 % fell by at least half

Coordinate axes

Eight axes and their observed lows

Each axis shows its high and low. The dot marks the low. Months with fewer than 1,000 messages are excluded from extrema selection because sparse denominators make comparisons unreliable.

Monthly human messages
Staffing changes and channel migration are not controlled. Compare patterns with context.
70.2 ×
High 2016-08 436,944 mentions Low 2026-02 6,225 mentions
Monthly active speakers
Accounts that posted in the month, not employee headcount.
12.1 ×
High 2017-01 266 speakers Low 2026-02 22 speakers
Top-ten speaker share
Fewer speakers alone can increase this share. Do not compare absolute values without context.
4 ×
High 2026-01 94.1% Low 2017-01 23.5%
Acknowledgement-only message share
Regular-expression pattern counts, not semantic judgments.
3 ×
High 2025-09 7.4% Low 2017-03 2.5%
Question-pattern share
Question marks and interrogative endings; this does not measure question quality.
1.7 ×
High 2025-10 15.4% Low 2016-02 9%
Night message share (22:00–06:00)
Digital activity timestamps are not direct evidence of sleep loss or burnout.
Zero observed Low: 0
High 2026-02 3% Low 2020-01 0%
Messages of at most 19 characters
Length is not a measure of information content or decision value.
1.3 ×
High 2016-03 79.1% Low 2025-12 59.1%
Replies within five minutes
Response needs and alternative channels such as calls or meetings are not controlled.
1.2 ×
High 2019-07 86.7% Low 2025-12 71.3%

How to read

A comparison, not a forecast

The sample is one company. This trajectory cannot estimate another company’s decline probability . It can describe observed depth : one case establishes that such a trajectory occurred.

The claim is not “you will fail”, but “Here was the observed low on this axis. Where are you now?” . The first claim cannot be validated here; the second can be challenged.

Each axis lists its confounders. The top-ten share can rise simply because speaker count falls; do not compare absolute values without context. Night activity is a timestamp observation, not direct evidence of sleep loss or burnout.