Cultural Entropy Mapping: The Second Law of Thermodynamics Applied to Engineering Orgs

60-Second Summary

Second Law of Thermodynamics: closed systems drift toward higher disorder.
Organisations are not closed — but they require continuous energy input to maintain low entropy.
Cultural entropy shows up as: undocumented knowledge, siloed teams, repeated incidents, knowledge concentrated in 'bus factor 1' people.
Map entropy quarterly using 6 signals; inject structure where it spikes.
Companies with formal entropy programs (Atlassian Team Playbook, Spotify Bot, Etsy Knowledge Maintenance) cut repeated incidents 40–60%.

Walk into a 5-year-old engineering org. Half the docs are out of date. Three teams use 'observability' to mean three different things. The senior staff engineer who knows the auth system is on holiday and nobody else can deploy. Nobody did anything wrong. The system just… drifted. That is entropy.

Entropy in 60 seconds

In thermodynamics: closed systems trend toward maximum disorder. In organisations: without continuous structural energy, knowledge fragments, processes decay, tribes form. The fix is not a one-time clean-up — it is a maintenance rhythm.

The 6 entropy signals

Signal	Low entropy	High entropy
Bus factor	≥ 3 per critical system	1 — one person knows it
Doc freshness	90% updated in last 6 months	30% updated in last 2 years
Repeated incidents	< 10% recurrence rate	> 30%
Cross-team contributors	≥ 15% of PRs cross teams	< 5% — full silo mode
Vocabulary drift	Shared terms across teams	Each team has its own dialect
Postmortem cadence	Every incident, within 7 days	Optional, frequently skipped

−47%

repeated incidents in orgs with formal knowledge maintenance

Etsy Engineering Blog, 2023

3.2x

MTTR for systems with bus factor 1 vs ≥ 3

PagerDuty State of DevOps, 2024

60%

of staff+ time at high-entropy orgs is spent on 'archaeology'

Stack Overflow 2024

What to inject to lower entropy

Maintenance time as a budget line — every team allocates 15–20% of capacity to docs, refactors, and ownership transfers.
Quarterly 'bus factor audit' — every critical system named, owners listed, single-owner systems get a buddy.
Mandatory blameless postmortems on every incident; track recurrence rate as a top org-level KPI.
Rotate engineers across teams quarterly at the staff+ level — keeps the social graph connected, breaks vocabulary drift.
Internal conferences and demo days as 'entropy fighters' — pumping shared context across the 50/150 layer (see the Dunbar article).
Doc freshness bots that flag pages not updated in N months — automation handles the boring half.

Entropy maintenance loop

Measure
Pull 6 entropy signals quarterly
→
Map
Heatmap by team — find the hotspots
→
Inject
Allocate maintenance time, rotate people, schedule postmortems
→
Re-measure
60 days later, look at the deltas

Why this matters now

AI-assisted coding is accelerating both the rate of shipping AND the rate of entropy. Code that nobody fully understands ships faster. Without entropy maintenance, the AI productivity gains are eaten by the compounding chaos within 18–24 months.

Takeaways

Orgs drift toward chaos by default. Stopping that is HR's job.
Entropy is measurable in 6 concrete signals.
Maintenance is a budget line, not a heroic act.
AI accelerates both shipping and entropy — the maintenance discipline matters more, not less.

References

Atlassian Team Playbook — Health Monitor — Atlassian
Etsy Engineering — Knowledge Maintenance — Code as Craft
PagerDuty State of Digital Operations 2024 — PagerDuty
Spotify Engineering — Bot maintenance — Spotify