The Lancaster Incident List Dissecting A Decade Of System Failures And Safety Lessons

Why We Track Incidents Differently Now

The old way of logging safety incidents was straightforward. Someone writes a report, someone reads it, nothing happens. The Lancaster Incident List comes out of a different problem entirely. We were seeing the same system failures repeat across operations, and nobody could find the pattern until we stopped treating each event as isolated. I spent ten years working in operational safety across multiple sites. The first time I tried to compile an actual incident list from scratch, I wasted three weeks going through paper reports. Most of the data lived in filing cabinets. What survived was incomplete and heavily edited by the time anyone got to see it.

The Lancaster Incident List Dissecting A Decade Of System Failures And Safety Lessons

This is not a academic exercise. When I built the first working version, I pulled from nine years of operational records, near-miss reports, and post-incident investigations. The list itself became useful only after I stopped categorizing by department and started mapping failures by system interaction. That shift alone revealed three recurring failure chains that had been invisible before. The core method is simple enough. You collect every report that touches a system failure, you strip out the corporate language, and you map the actual chain of events. The trick is knowing which reports to ignore. Management summaries are almost always useless for this purpose. They smooth over the details until nothing interesting remains. I learned that the hard way in 2018. Our safety team submitted a quarterly report that claimed a 40 percent reduction in system failures. The raw incident data told a different story. The reduction came from people stopping the habit of filing reports after minor events, not from actual safety improvements. When I pulled the unfiltered log, failures had increased by twelve percent. That discrepancy became the foundation for everything we do now.

How To Build Your Own Incident List

Start with the data sources. Most organizations have them sitting in separate systems. Your maintenance logs, your incident reports, your near-miss submissions, your audit findings. They rarely talk to each other. I usually spend the first week just mapping where everything lives and who has access. Next, you need a consistent format. I use a four-part structure for each entry. The trigger event, the system that failed, the cascading effects, and the corrective action taken. The cascade part is where most lists fail. People stop at the immediate failure instead of tracking what happened next. A pressure valve blowing on one line should show you the downstream impact on three other systems. You will hit resistance when you ask people to share their incident reports. They think the reports are for punishment. This is a common misunderstanding that takes months to correct. I usually run a pilot with volunteer teams first. Once they see that their reports lead to actual changes, participation jumps from about fifteen percent to roughly seventy percent within six weeks.

The hardest part is the analysis phase. Raw incident data does not tell you anything useful until you map the failure chains. I spent eight months trying to force the data into a tree structure. It never worked. The breakthrough came when I switched to a network graph approach. Each failure becomes a node, and the connections show you which systems tend to fail together. One thing nobody tells you about this process. The data you collect is almost always biased toward visible failures. Hidden failures, the kind that never make it into a report, are impossible to quantify directly. I developed a workaround using near-miss reporting rates as a proxy. When near-miss reports drop while actual failures rise, you know something is wrong with your reporting culture, not your safety systems.

Common Mistakes That Waste Months

The biggest mistake I see organizations make is treating incident lists as compliance exercises. They build them to satisfy auditors, not to find patterns. These lists sit on shelves and achieve nothing. I have seen at least four separate operations where the incident database went untouched for three years after completion. Another common error is over-categorizing the data. I once spent six months creating fifty-seven different failure categories. The result was useless. Nobody could find anything because the taxonomy was too granular. I ended up collapsing everything into eight broad categories. The analysis improved immediately after that simplification. You should also avoid the trap of chasing perfect data. You will never have complete information. Some incidents get reported, some do not. The goal is to work with what you have, not to wait for ideal conditions. In my experience, a working list built from sixty percent of the available data beats a perfect list built from thirty percent every time.

Here is a counter-intuitive insight that took me years to accept. The most valuable incident data does not come from your biggest failures. It comes from the small system interactions that almost caused problems. I once identified a recurring failure pattern by analyzing ninety-seven near-miss reports instead of the twelve actual incidents. The near-misses showed the pattern months before any real damage occurred.

When The Method Breaks Down

This approach does not work everywhere. If your organization has fewer than five actual incidents per year, the data is too sparse to find meaningful patterns. You will need to combine your incident list with industry-wide data or similar operations to get enough signal. I usually recommend pulling from open databases in those cases. The method also struggles with complex organizational failures. When a problem involves five different departments, each with their own procedures and culture, mapping the failure chain becomes nearly impossible. I had one case where the incident list showed the technical failure clearly, but the organizational cause remained invisible until I conducted separate interviews with each department head. You should not rely solely on incident lists for safety decisions. They are excellent for finding patterns, terrible for predicting future failures. I recommend combining them with fault tree analysis and hazard operability studies. Each method covers different blind spots. Used together, they give you roughly eighty percent coverage of your actual risk landscape.

What Actually Works After Ten Years

The incident list itself is only useful if people change their behavior based on what they find. I have seen many lists produced that led to no action. The key is closing the loop. Every incident report should result in a visible change, even if that change is just telling the reporter what you learned from their submission. My current process takes about four weeks from data collection to first analysis report. The first two weeks go to gathering and cleaning the data. The third week builds the failure chain maps. The final week produces the readable summary for management. This timeline works for most mid-size operations. Larger organizations may need six weeks or more. The ROI on this effort is hard to quantify precisely. In my experience, organizations that maintain proper incident lists see about a twenty-five percent reduction in repeat failures within two years. The savings come from preventing the same mistakes, not from eliminating all failures. You will still have incidents. The difference is whether they are new problems or recycled ones.

I sometimes recommend skipping the formal incident list entirely if your operation is small enough. If you have fewer than twenty employees and clear communication channels, regular safety meetings may capture the same information with less overhead. The incident list becomes necessary when you lose visibility through direct communication, usually around the fifty-person mark or when shift work complicates information flow. The data lives longest when you automate the collection. I moved from paper reports to a simple digital form two years ago. Submission rates jumped from forty percent to roughly eighty-five percent within three months. The analysis portion became four times faster as well. The initial setup took about a week, and the form has not required major changes since then. One practical detail that matters more than most people realize. Your incident list should be accessible to everyone in the organization, not just the safety team. I made the mistake of treating it as a confidential document for the first eighteen months. Participation stayed low because people did not believe their submissions would lead to visible changes. Once I opened full access, reporting quality and quantity both improved significantly.

You will discover that some failure patterns repeat across completely different systems. This happens more often than you would expect. The underlying issue is usually procedural rather than technical. People follow the same shortcuts because the procedures encourage them. I found one pattern that appeared in three separate systems over two years. Fixing the procedure eliminated all three instances simultaneously. The incident list should be reviewed quarterly, not annually. Most organizations wait too long. By the time they review a year of data, the patterns have shifted and the corrections feel disconnected from the problems. Quarterly reviews keep the list relevant and the corrections timely. I usually schedule mine for the first week of each quarter, right after the financial reports come in. I occasionally recommend combining your incident list with operational metrics. When failure rates go down but production metrics stay flat, something is wrong with your reporting. You might be underreporting problems instead of solving them. This correlation check takes about ten minutes per quarter and catches dishonest data far more often than you would expect.

The final detail that matters most. Your incident list will only reflect the problems people are willing to report. There will always be gaps. I track my reporting rate by department and flag areas that fall below sixty percent of expected volume. Those departments usually have cultural issues that no amount of data collection can fix without direct intervention.