The Slow Creep
Sidney Dekker, a human factors researcher and former airline pilot, coined the term Drift into Failure to describe how organizations slide toward catastrophe not through sudden negligence, but through a long series of small, rational, successful decisions.
Disasters rarely happen because something suddenly breaks. They happen because an organization slowly, incrementally drifts toward the boundary of safe operation — and never notices, because nothing bad happens along the way.
The Drift Mechanism
The process follows a predictable pattern:
- Optimization pressure: Management pushes for faster, cheaper, better. Deadlines tighten. Budgets shrink. Headcount is frozen.
- Incremental trade-offs: Engineers make small, local adjustments to meet these goals. "We'll skip the canary deployment today to hit the release window." "We'll reduce the replica count to save cloud spend."
- Reinforcement by success: The risky behavior works. Nothing breaks. The team ships on time. The cost savings are celebrated. The trade-off is repeated.
- Normalization of deviance: The risky behavior becomes the new standard operating procedure. "We haven't run canaries in six months and we're fine." The original safety margin is forgotten — or dismissed as unnecessary overhead.
- The accident: Eventually, the system crosses an invisible safety boundary. The same trade-off that worked 47 times fails on the 48th, under slightly different conditions.
The Crucial Insight
The decisions that led to the failure looked like rational, successful choices at the time they were made. The engineers were not incompetent. The managers were not malicious. The failure is the product of success — of optimization working so well, for so long, that the safety margins were consumed without anyone noticing.
This is why post-mortems that conclude "the team should have followed the process" miss the point entirely. The team stopped following the process because following it appeared unnecessary. That appearance was the drift.