Ask most heads of safety what happens after an incident, and you'll hear a familiar pattern. An investigation opens, an owner is assigned, findings get written up, a report gets filed. Ask what happens to those findings six months and two incidents later, and the answer is usually: not much. They sit in a different system from the risk assessments that predicted the same failure modes, disconnected and easy to lose.
Most investigations also stop at the front line. A worker made the wrong call, so the worker is the cause, case closed. That's rarely the full picture. The conditions that set someone up to make that call, training gaps, staffing decisions, workload, unclear standards, sit further back in the organization. If an investigation doesn't reach them, they stay in place for the next incident too.
What is HFACS?
The Human Factors Analysis and Classification System (HFACS) examines human error by classifying it against a structured taxonomy. Developed for aviation, it's now used across industrial domains. The premise is simple: in complex socio-technical systems, an accident is rarely the result of one person's mistake. It results from multiple contributions, some immediate, some sitting further back in the organization, from unsafe acts at the front line up through preconditions, line management deficiencies, and fallible decisions at the organizational level.
Seeing it in practice: Colgan Air Flight 3407
On the night of 12 February 2009, Colgan Air Flight 3407 was flying from Newark to Buffalo with 49 people on board. The crew violated the Sterile Cockpit Rule during taxiing and continued talking through the flight. The first officer calculated approach speed for non-icing conditions when icing-adjusted speed was required, and the crew failed to monitor a known risky maneuver closely enough. When the stick shaker activated, the captain raised the angle of attack instead of lowering it, then overrode the stick pusher, the last functional barrier. The aircraft stalled, entered Loss of Control In-flight, and crashed 27 seconds later. All 49 on board and one person on the ground died.
Why this matters beyond one flight
Run through BowTieXP and IncidentXP, this analysis isn't a one-off report. The same barriers modeled proactively in a bowtie become the backbone of the reactive investigation, so nothing is rebuilt from scratch. Classified causes like fatigue, inadequate training, and culture feed into Safety Performance Indicators that an organization can track across its entire operation, not just after something goes wrong.
The lesson from Colgan Air isn't that the crew didn't make mistakes. It's that stopping the analysis there would have missed what really needed to change.
Read the full expert insight document
The full Expert Insights piece by Captain Augusto Claudio Derghi walks through the complete causation paths, from unsafe acts up to fallible decisions, and shows exactly how BowTieXP and IncidentXP surface each layer.