Run a failure through a connected system and a good model shows you two things: the systems significantly affected, listed with scores, and the lines that show how the failure travelled to reach them. It is a clean picture, and most days it is the one you want. Then on some systems you notice connecting lines running out to faint nodes at the edge — nodes that are not in the results list. The lines appear to go nowhere.
They are not a glitch
Those faint systems were genuinely reached. They absorbed a small shock and passed it on. The model is showing you something real — it just is not telling you what it is. Behind the picture sits a single threshold: absorb more than five percent of the original shock and you are reported, bright and named; absorb less and you are reached but not significantly affected. Everything above the line becomes part of the story; everything below it becomes background, present in the picture but absent from the analysis.
A visible mystery is the worst option
There are two honest things to do with the effects below the line. You can decide they genuinely do not warrant reporting, remove the faint lines, and show only what made the cut — accepting that you are trading weak-signal information for a clean presentation. Or you can decide they carry structural insight and give them a home: keep the lines, and add a lower tier that says plainly these systems were reached, here is how far the failure travelled, here is what it touched on the way out. What you should not do is what most tools do by default — show the lines and stay silent. That raises a question it refuses to answer, and leaves the analyst to guess. A clear omission is better than a visible mystery.
One picture cannot hold it all
The same tension shows up in static reporting. The Chronic Risks Analysis mapped its relationships across twenty-six separate diagrams, each able to show only a fraction of the connections — so a link that exists on one page is missing from another, not by error but for want of space. Each diagram is an honest partial view; the full network appears on none of them. Readers who want the whole picture have to cross-reference all twenty-six and infer the rest. Most won't.
So what?
The goal of visualising risk is not a prettier chart — it is a picture that tells the truth about its own limits. Whether a connection matters, where a faint line actually leads, what a threshold quietly removed: these are choices, and a good visual surfaces them instead of hiding them. Not a better report — a model you can explore, where the lines that go nowhere become a question you can actually follow.