A budget stress test runs a plan thousands of times with every driver moving inside a range the owner would accept, and returns three things the single number hides: the chance of hitting it, the downside, and the two or three drivers that decide the year.
A budget is one number per line. A budget stress test treats each driver behind those lines as a range instead of a point: volume between −12% and +8% of plan, steel between −15% and +25%, and so on. It then runs the plan 10,000 times, drawing every driver from its range with the links between them intact, and counts the outcomes. The result is a distribution, not a cell: the share of runs that reach the budget, the median, the 1-in-10 downside, and the ranking of drivers by how much of the swing they carry.
Four steps. First, the plan is laid out as a driver model: revenue is volume times price, steel cost is a price index times volume times the plan amount, and so on down to EBITDA and free cash flow. Second, each driver gets a low, a high and a shape (uniform, triangular or normal), set by the person who owns it; correlated drivers, such as steel and components, are linked. Third, a Monte Carlo engine runs 10,000 scenarios through the model with a fixed seed, so a changed range changes the answer and nothing else. Fourth, the outcomes are read back: the odds, the percentiles, a tornado of drivers moved one at a time, and goal-seek questions such as "at what steel price is free cash flow zero".
The chance of hitting the budget, as a percentage. The median outcome and the P10 and P90, so the board sees the band, not only the point. The drivers that carry the swing, ranked, with the share of the total each one holds. What-ifs on the same scenarios: narrow one range and see the odds move. And a goal-seek: the value of one driver at which a line reaches a target, with everything else at plan.
Sensitivity analysis moves one driver at a time and shows a slope; it says nothing about how likely the move is or what happens when several drivers move together. Scenario planning picks three stories, best, base and worst, and attaches no probability to any of them. A stress test moves every driver at once, inside ranges with odds, thousands of times. It keeps both tools: the tornado is the sensitivity view, and any single run can be read as a scenario, but every answer now comes with a probability.
At ScenarioCast the first run is the Kick-start: ten working days from the budget file to the board one-pager, $20,000 fixed in USD, up to 25 drivers, EBITDA through to cash flow, covenant headroom, built fully for you. If the driver model does not reconcile to your budget to the dollar, you do not pay. After that the Studio re-runs it whenever an assumption moves: $1,000 a month self-managed, $6,000 a month managed. The model stays an Excel file; nothing is migrated.
The public use case runs the full Studio on a US manufacturer's cash-flow plan built from public filings, with no login. The articles apply the method to BASF, Airbus, ArcelorMittal, a tariff case and a $400M US manufacturer. Or book thirty minutes and bring your own file.
Thirty minutes on Microsoft Teams, every weekday, in your own time zone.