The forecast was never the problem. The paradigm underneath it is.
A decade of better predictions, better visibility, and better firefighting has not made supply chains more resilient, because the problem was never the forecast. It is what planning does with it: reduce an uncertain business to single numbers, optimize the average, and commit the enterprise to a future that arrives, at best, occasionally.
Four assumptions the planning paradigm still runs on. All four have expired.
The same reduction happens everywhere. Demand becomes one number per item per period. A supplier who delivers in 20 days or 45 days becomes “30.” A line that yields 96% or 78% becomes “90.” Each simplification is individually reasonable. Compounded across a network, they produce a plan for a company that does not exist, and the enterprise spends real money defending it: buffers sized for the wrong world, service misses in the good weeks, stranded capacity in the bad ones. Averages are useful descriptions. They are dangerous commitments when the variability around them determines the economics.
Sequential S&OP settles demand first and makes everyone downstream pay for it.
Demand commits, supply reacts, inventory fills the gaps. Every handoff strips out uncertainty the next function needed to know about, and the cost surfaces two functions later where nobody can trace it.
Silo optimization wins every battle and loses the P&L.
Each function optimizes its own number. The enterprise absorbs the interactions: the cheap procurement decision that starves the line, the service target that quietly costs three points of margin.
Single-future plans bet the quarter on one scenario.
Reality arrives as a distribution. A plan that only considered one future has no answer for the others except the war room.
Stale master data means the plan describes a company from three years ago.
Parameters drift from the reality they once described, and nobody owns recalibrating ten thousand of them by hand. The plan inherits every stale number as a fact.
A better forecast does not retire a single one of these. A different decision architecture retires all four.
A proposal is not a decision.
A planning system produces a recommended plan. That is a proposal. It carries no account of the futures in which it fails, no floor it has promised to hold, no record of the alternatives it set aside, and no authority to act when reality drifts. Someone in a meeting still has to turn it into a decision, and the part that makes it a decision happens entirely outside the system that produced the proposal.
A governed decision is a different object. It states the floors it must hold, service, margin, cash, as explicit constraints, not preferences. It is scored against the range of futures it might meet, so its resilience is a number, not a hope. It records what it chose and what it rejected, so it can be audited and improved. And it is bounded by a Decision Policy: a first-class artifact that says what may be decided automatically, within which limits, and when a human must be in the loop.
The gap between a proposal and a decision is exactly the work the enterprise currently does by hand, every cycle, in the room. VYAN closes that gap by making the decision, not the plan, the thing the system produces.
A recommended plan tells you what to do. A governed decision can defend why, hold a floor while it does, and act without convening a meeting.
A System of Intelligence, above the System of Planning and the System of Record.
The enterprise already runs two stacks. The lowest holds the truth of what is. The middle proposes what could be done. Neither was built to govern a decision under uncertainty, so VYAN adds a third layer above them, not beside them. It does not replace your ERP or your planning suite; it sits on top and turns their outputs into governed, scored, defensible decisions.
Where decisions are made, governed, and scored. It reads the record and the plan, applies the Decision Policy, prices the tail into the objective, and emits a decision with its resilience attached. This is the layer that was missing, and it is the layer VYAN is.
Where proposals are generated: the planning suites and optimizers that turn demand and supply signals into a recommended plan. Useful, but it stops at the proposal. It cannot tell you which of its plans survives contact with the futures it didn't plan for.
Where the facts live: the ERP and transactional systems that hold orders, inventory, and master data. Authoritative about the past and present, silent about what to do next, and only as current as its last calibration.
How the SOI does this, the math, the substrate, the Decision Policy object, is the Platform. This page is the argument for why the layer has to exist at all.
A·I·R, and Resilient is the point you can act on today.
The System of Intelligence operates along three properties: Autonomous, Integrated, Resilient. They compose into one posture, but they do not arrive at once. Resilient is what we put in your hands today: the decision that holds across the futures it might actually meet. Autonomous and Integrated are the road ahead, and they have their own sessions in this series, so the examples below land where the value is now while pointing at where it goes next.
Committing to a plan whose P50 lead time already exceeds the ERP's number.
The ERP carries one lead-time number for the ocean lane. The actual lane is a distribution, and lately its median has crept past the number the plan still trusts, so the “on-time” plan is already late at the fiftieth percentile, before anything goes wrong. Nobody sees it, because the average on the dashboard hides the days that cost money.
A resilient decision reads the lane as a shape, not a constant. It prices the expedite, the ocean-to-air switch, against the probability curve of the tail, and decides the cheaper-on-average path is the more expensive one once you weight the days you actually pay for. The expedite stops being a panicked Tuesday call and becomes a priced, defensible decision made before the breach.
A demand pull-in that quietly breaks a margin floor in another function.
Commercial pulls an order forward to win the quarter. The plan accepts it: the fill rate looks great, the demand number is met. What the plan cannot see is that serving the pull-in means expediting freight and breaking into premium capacity, and somewhere two functions away, that quietly pushes the order below its margin floor. Each silo's slide is green; the enterprise just lost money it will only find at quarter close.
An Integrated decision is cross-functional by construction: not ERP-to-ERP plumbing, but the demand, supply, freight, and margin constraints solved as one objective. The margin floor is a constraint the decision must hold, so the pull-in is either re-shaped to keep the floor or surfaced as the trade-off it actually is. That is the next half of the argument, in its own session.
Plans break. Policies hold.
A plan is a single-future bet. It picks one version of next quarter and optimizes for it, and the mean it optimizes against hides the expensive days: the congested port, the supplier miss, the pull-in that breaks a floor two functions away. The average looks fine right up until the tail arrives, and the tail is where the money is lost. This is the distribution the whole argument turns on: an enterprise does not live at its mean, it lives across the full spread, and the cost of the spread is real even when the average is calm.
A policy is the opposite kind of object. It does not bet on one future. It states the floors that must hold and then holds them across the hundreds of futures the enterprise might actually meet. Service floor, margin floor, cash floor: a policy keeps them standing whichever future arrives, and scores every decision by the share of futures in which it remains the right call. A plan optimizes for the future it guessed. A policy is built so the guess never has to be right.
Plans break, because they bet on one future. Policies hold, because they were built for all of them.
If this is the right diagnosis for your enterprise, the next step is to see your own numbers under it. PULSE is a short, guided diagnostic that scores where your decisions break today and what a resilient policy would hold instead, grounded in your public financials, no fee, no deck.
Want the argument in plain terms first? SOI / SOP / SOR and Resilient Decisions in Learn.
Built from three decades inside global planning transformations.
VYAN founder Ashutosh Bansal led supply chain practices at PwC, IBM, and SAP, with transformation work spanning global enterprises in high tech, consumer, industrial, and energy. VYAN exists because the paradigm he helped scale is no longer the right one.
Meet the founder →