Most risk registers sort by the two things nobody can know.
Likelihood and impact are both guesses about a future you do not have yet. This framework sorts by two properties you can check today: how the payoff is wired, and how the tail behaves. Those two answers put a risk in one of five rooms, and the room tells you what to do.
Start the quiz to find your quadrant.
Two questions, about thirty seconds
Question 1 of 2
Does the damage stop where it starts?
Question 2 of 2
Could one event be far bigger than anything in your records?
One more, because fraud sits across all four
Is any of that true only because someone says it is?
Documented, foreseeable risks that were ignored anyway. The failure is attention, not information. Contained and predictable does not mean harmless. This is the room where the hazard was already written down and nothing about it forced anyone's hand.
The tell. Look for the paper trail that predates the event. In this room it almost always exists.
What to do
- Go and read what has already been written down. Here the finding is usually filed before the failure.
- Give every open warning a named owner and a date. Warnings that belong to everyone belong to no one.
- Check what changed after the last inspection. If nothing changed, the finding never landed.
- Ask who is paid to say stop, and whether saying it has ever cost them anything.
The rare, high-impact event that was always coming, and was always going to be called unforeseeable afterwards.
The tell. Ask what the defences were sized against, then ask what the worst case on record actually is. The gap between those two is this room.
What to do
- Decide the response before the event and write it down. There is no time to reason once it lands.
- Size the defence against the worst case on record, not the worst case anyone remembers.
- Choose deliberately between staying out of the way and building a defence, and price what avoidance costs.
- Insure what you cannot cap, then check the insurer could pay and read what the policy excludes.
Tightly coupled systems where one small fault cascades and takes down the whole machine.
The tell. Look at what cannot be switched off independently. That is where the cascade will run.
What to do
- Build spare capacity into anything that cannot be switched off on its own.
- Test the backup on a schedule. A backup nobody has restored from is an assumption.
- Rehearse the response while nothing is wrong. Mid-cascade there is no time to improvise.
- Keep the boring maintenance current. Expired certificates and unpatched software start most of these.
The world stops matching the model. Regime change and leverage turn a small error fatal.
The tell. Ask what the position assumes is stable, then ask what happens on the day it is not.
What to do
- Leave. This is the one room whose correct answer is to be in a different room.
- Cap the downside in the contract, so the tail cannot reach past a number you chose.
- Cut the borrowing. Leverage is what turns being wrong into being finished.
- Stop refitting the model. It has failed structurally, and another calibration buys confidence, not safety.
The fifth quadrant, where the thing was never real. The tell is that the story is too clean.
The tell. Look for the number that is too good for the world that produced it, then ask who would know if it were not true.
What to do
- Verify outside the house. Friendly auditors, internal valuations and uninspectable quality control are the pattern.
- Treat urgency and exclusivity as findings. Both exist to shorten the time you spend looking.
- Name who would know if the number were false, then ask what makes them independent.
- Take a story that is too clean as the warning, not the reassurance. Here the job is detection.