Lessons Learned – How to Make Them Stick

Write to: jc@knowledge1.net. 

Incidents that happened before happen again, showing that we either are not paying attention to incidents or are not learning the lessons from those incidents.  On October 1989, Phillips 66 experienced a major explosion at its Pasadena, Texas, facility which killed 23 people and left more than a 100 injured.  This explosion precipitated the enactment of OSHA’s PSM Standard, 29 CFR 1910.119.  Among the causes of the incident, OSHA cited faulty maintenance procedures.  Ten years later, in 1999, an explosion rocked the same Pasadena facility resulting in two fatalities.  And a year later (2000) another explosion occurred in the same facility resulting in one fatality and 71 injured.  Although the immediate causes for all these explosions were different, the common thread was maintenance operations.  In another case, at Hoeganaes Corporation there wasn’t even a significant time span before incidents were repeated.  In 2011 the company experienced three incidents with a total of 4 fatalities and one serious injury.  All three incidents were related to a dust fire or explosion.  A recent editorial in Process Safety Progress shows examples and reiterates that we don’t seem to learn from past history.

Why Are Lessons Not Learned?

Lessons are not learned, not because of lack of trying.  The Global Congress on Process Safety (AIChE) which meets once a year, has a joint session of all its tracks dedicated to lessons learned.  CCPS has published books and maintains a database of incidents for participating companies.  In the last twenty years 297 articles on lessons learned have been published in Process Safety Progress, and articles continue to be published elsewhere.  A monthly bulletin from AIChE provides an account of an incident or near miss from which we can learn.  The US Chemical Safety Board (CSB) publishes online reports of its investigations and offers videos that demonstrate how the incident happened.

Companies typically circulate to employees the results of their own incident investigations and the reports are usually available internally.  But the lessons from those incidents may get lost, maybe because they don’t have immediate impact, or because they get forgotten when they should be applied, during process design or a process hazards analysis (PHA).

In essence, in spite of the wealth of incident information, the task of distilling a lesson from an incident, communicating it, and having people remember it when it’s needed, is a difficult task.  Typically, in the long term we retain about 10% of the information we receive during training.  And training is much more than sharing information where retention may be less than 2%.

How to Make Lessons Learned Stick

In order to really learn a lesson, the lesson needs to be delivered in a form that will be unforgettable, it needs to be woven into the fabric of the company, that is, translated into everyday use, and it needs to be periodically reinforced.

The Delivery of the Lesson

Studies have shown that long term retention of knowledge depends on the method of delivery of the training.  The amount retained can vary from 2% to 90% depending on whether the information was transmitted by reading, by oral presentation, visually, and/or by immediate application (the actual numbers have not been substantiated).  Presenting the information in an impactful visual method will lead to be retained longer and the videos from the CSB are an excellent example.  In Figure 1, taken from one of the videos, we can see two workers that have collapsed and died inside a reactor that had been inerted with nitrogen.  The workers had finished the job when they discovered that a tape had been left inside the reactor.  One of the workers tried to fish the tape with a wire while sitting in the open manhole.  He was overcome by the nitrogen and fell into the reactor where he collapsed from the nitrogen atmosphere.  When seeing this, his fellow worker tried to rescue him by inserting a ladder into the reactor and lowering himself into it.  He also collapsed.  This was noticed by a supervisor, but by the time that the proper equipment had been obtained for a rescue, the workers had died.  Showing the video will have much more impact than giving a lecture.  It is very important, though, to provide the correct lessons learned from the incident, adhering to the top two or three.  In this case they would be that (1) nitrogen is an insidious killer, and (2) do not attempt to rescue a fellow worker without first ensuring that you have the proper equipment to do so.

Figure 1.  Collapse of two workers in a nitrogen atmosphere (from the CSB video).

Integration into the Company’s memory  

If the lesson is well delivered but not applied, it may stay in people’s memory but it won’t be practiced and the value of the lesson will be lost.  Thus, the lesson needs to be integrated into the practices of the company.  A repository to capture the lesson is necessary and its essence applied to the company’s technical, operational and safety procedures.  

Reinforcing the Learnings

In spite of everything, memory fades with time.  Also, there’s turnover in the plant, new technologies are adopted, and procedures evolve.  There needs to be periodic reinforcement of the lessons learned in terms of the rationale for doing things the way there are being done.  Two ways of doing this reinforcement are:


By using the methods described here, lessons are learned through visualization, constant application and reinforcement.  Since the lessons become part of the knowledge of the company, and are integrated into its everyday activities which should include periodic refreshing, these lessons will stick.

A full article on the subject appears in: Chosnek J. Lessons learned—How to make them stick. Proc Safety Prog. 2020; e12198, https://doi.org/10.1002/prs.12198.

The pre-peer reviewed paper can be downloaded at http://knowledge1.net/publications/

In order to comply with the requirements of MOC, whether because of regulations or because it makes good process safety sense, companies have set up elaborate systems that sometimes defeat the purpose.  The purpose of MOC is first, recognize that there is a change, and then determine if that change is safe.  Thus, a safety analysis is always required.  Should it always be a detailed safety analysis?  A full PHA with a team that includes a facilitator and scribe?

The answer is “no”.  The safety analysis should be commensurate with the complexity of the change.  Doing a detailed safety analysis every time takes away resources that could be used on other operations that have a higher risk, and that would make them safer.  For example, if you are going to change the driver of a pump to one with higher horsepower, the safety analysis could be made by a small group consisting of an experienced process engineer and a pump expert.  If the safety analysis shows a potential for significant impact on the process, then a more detailed study would be required.  On the other hand, if you are going to change the capacity, size or type of a process heat exchanger, the potential impacts are so many, that a detailed analysis should be performed.

The important aspect of a safety analysis is a factor that tends to be glossed over: the quality of the analysis.  I have reviewed dozens and a great many are left wanting. Review of the study by a technically competent member of the facility, that is not involved in the change (and especially doesn’t approve the change), goes a long way towards improving the quality of the safety analysis. This is what’s called peer review.

One last factor that typically doesn’t get included in the safety analysis is review of previous analyses.  Do you review your last 5-year PHA to identify potential impacts of the change?  Remember that a very detailed analysis was done at that time.  Do you review your past MOCs to see if you are modifying a past modification?  Are you unwittingly affecting a layer of protection that was put in place (and justified) in a past MOC?

As all these are time consuming activities, it pays to have an electronic PHA and an electronic MOC system.  Just make sure that they are easily searchable.