The two numbers in the Fatigue and Risk Index, and how rail gets them wrong
The first time I ran a roster through the HSE Fatigue and Risk Index, I got two numbers back and confidently misread both of them.
One was a 22. The other was a 1.3. I assumed they were two flavours of the same thing, glanced at the threshold figures everyone quotes, decided the pattern passed, and moved on. It took an evening with the ORR's appendix on assessment tools, which is drier than the Gobi but genuinely worth reading, to understand that I had treated a probability and a relative risk as interchangeable, and treated an indicative flag as a pass mark. Both mistakes are everywhere in this industry. This post is the explanation I wish someone had handed me first.
Two indices, not one
The FRI produces two separate numbers for any working pattern, and they measure different things.
The fatigue index is a probability, expressed as a percentage, that someone working the pattern will feel very fatigued at some point during the shift. A fatigue index of 20 means roughly one person in five on that pattern is likely to hit high sleepiness. It is about how people will feel.
The risk index is relative, not absolute. It estimates the risk of a fatigue-related incident compared to a reference pattern: twelve-hour shifts on a two-day, two-night, four-off cycle. A risk index of 1.4 means an estimated 40 percent more incident risk than that reference. It is about what might happen.
Here is the detail that finally made it click for me: the two indices peak at different times of night. Risk peaks close to midnight. Fatigue peaks around five hours later, in the small hours before dawn. They are modelling different phenomena, which is why a roster can look acceptable on one and poor on the other. A string of short early starts can feel brutal, high fatigue index, while carrying modest incident risk. A pattern with one long night at the wrong point can do the reverse. If you only track one number, you are seeing half the picture and you do not get to choose which half.
What a threshold is, and is not
The thresholds circulating in the rail industry came from Health and Safety Laboratory work commissioned after the industry adopted the FRI. They were described from the start as indicative: a way of flagging patterns that deserve attention, not a pass mark.
I understand exactly why they harden into pass marks anyway. I have sat in enough delivery meetings to know what happens to a nuanced number once it enters a governance pack. Someone needs the roster signed off by Friday, the appendix nobody read said 35, the roster scores 34, and the conversation is over. The number was built to start conversations and it gets used to end them.
But a roster scoring 34.9 is not safe and one scoring 35.1 is not dangerous. The tool's own guidance says there is no ideal risk, only lower and higher, and that each operation should agree what is tolerable given the nature of the work. A shunter and a signaller with the same score are not carrying the same consequence if fatigue wins. The regulator's own commentary is blunt about this: the tool compares the relative merits of working patterns. It does not issue verdicts.
The gap the numbers cannot see
The FRI scores the roster as planned. It knows nothing about the overtime swap agreed by text on Sunday night, the two-hour commute, the second job, or the phone that rang at 3am. Every published evaluation of the tool makes the same observation: real fatigue exposure is almost always worse than the planned roster suggests.
Before I built software I ran a coffee shop, and I wrote the rota every week. I can tell you with certainty that the rota on the wall and the week people actually worked were never the same document. Swaps, cover, someone's sick kid. That was espresso machines and nobody got hurt. In rail, the gap between the published roster and the worked one is exactly where the risk lives, and a once-a-year assessment at roster design is an audit of a fiction.
The practical response is not to abandon the tool. It is to score what actually got worked, not just what was published. Continuously, not annually.
Where this leaves you
Three habits separate operators who use the FRI well from those who merely run it:
- Score both indices, always. Report them side by side and explain the difference to anyone reading them. If your board pack shows one number, ask which one and why.
- Treat thresholds as tripwires, not gates. A breach starts a conversation about the pattern, the individual, and the task. It does not end one.
- Re-score against actuals. The planned roster is the hazard assessment. The worked roster is the exposure record. You need both, and the second one is the one that will appear in the investigation report if it ever comes to that.
RotaPulse exists because I wanted the third habit to be cheap, and because the standard tool for the job no longer officially exists. It scores rotas against the HSE FRI in the browser, as planned and as worked, and shows the danger before the incident rather than in the report after it.
Try it on one of your own rosters with the free calculator. No signup, no spreadsheet, no macros security warning.
Sources
- ORR, Managing Rail Staff Fatigue, Appendix D: Benefits and limitations of fatigue risk assessment tools
- HSE Research Report RR446, The development of a fatigue / risk index for shiftworkers (Spencer, Robertson and Folkard, 2006)
- HSE, Human factors: fatigue
- RSSB, Fatigue Risk Management Systems
- Health and Safety Laboratory (2008), Evaluation of the UK Rail Sector Initial Fatigue and Risk Index Thresholds.
- Fatigue and Risk Index Calculator Version 2.2 User Guidance (HSE), mirrored copy