Two engines: diagnose, then decide.
A resilient decision is never one step. First you map the range of what could happen. Then you commit a single move that holds across all of it.
The whole category fits in one distinction. Making a decision under uncertainty is two jobs, not one, and conflating them is where most tools fail. VYAN keeps them separate and names them: Risk Diagnostics asks what could happen? and Resilience Optimization asks what should we do?
RDA · Risk Diagnostics — what could happen?
The first engine is diagnostic. It refuses to collapse the future into a single guess. Demand isn't one number — it's a range with a likely middle and tails on both sides. Lead time isn't one number — it's a spread of plausible durations, most of them ordinary and a few of them painful. Yield, cost, capacity: all ranges, never points.
Risk Diagnostics takes those ranges and works out what they imply together. It maps the spectrum of futures the business could actually live through — the gentle ones, the ordinary ones, and the expensive ones where several things go wrong at once. The output is not a prediction. It is a clear-eyed picture of the distribution of outcomes: here is what's likely, here is what's possible, and here is where it hurts.
A map, not a bet. It shows the whole spread of what could happen — and, crucially, where the costly tail lives — so the next decision is made with the downside in full view rather than discovered after the fact.
ROA · Resilience Optimization — what should we do?
The second engine is decisive. Given the spectrum the diagnosis laid out, it has to commit to a single move — because you can only place one order, set one buffer, make one allocation. It cannot hedge by running a different plan for every future. It has to pick one.
The trick is which one. A naive optimizer picks the plan that scores best on the most likely future and ignores the rest — which is exactly the brittle move that breaks the moment reality drifts. Resilience Optimization does something harder: it evaluates a candidate plan against the entire spectrum at once, and chooses the one that holds up best across all of it, with explicit weight on not breaking in the expensive tail.
One plan, scored against thousands of futures — and chosen because it survives the bad ones, not because it wins the lucky one.
Why the split matters
Keeping diagnosis and decision separate is what makes the whole thing honest. If you let the optimizer also decide what the future looks like, it will quietly pick the future that makes its preferred plan look good. Splitting the two forces an unbiased map first, then an accountable choice against that map. You can inspect the diagnosis on its own terms, and you can inspect the decision on its own terms.
It also matches how good operators actually think. They don't skip straight to an answer; they first ask "what are we up against here?" and only then ask "so what do we do?" The two engines make that discipline systematic, repeatable, and fast — turning a process that used to take weeks of meetings into something the layer does continuously.
RDA maps the range. ROA commits the move. Diagnosis without decision is just anxiety; decision without diagnosis is just a guess. Resilience needs both.