Re-forecasting: when to change the plan, and when to hold your nerve

Forecasting · ~6 minute read

The plan is wrong — now what?

Every forecast is wrong; the only question is by how much and what you do about it. A week in, actuals are running above plan. Do you re-forecast and re-plan around the new reality, or hold your nerve and let it settle? Both answers can be right, and both can be expensive. Re-forecast too eagerly and you churn the plan on noise, exhausting the operation with constant change and chasing a signal that wasn’t there. Hold too long and you defend a number everyone can see is broken, losing the chance to react while there was still time. Knowing which is which is one of the quietly hardest judgements in planning.

Signal versus noise, and time to act

Two questions decide it. First, is the variance signal or noise? A single odd day is almost always noise; a consistent bias in the same direction over a meaningful run is signal. Forecasts are supposed to be wrong day to day around the right average — that random scatter is not a reason to re-forecast. A persistent lean — every day above plan, the gap widening — is. Second, is there still time to act? A re-forecast you can’t resource is just paperwork; the value of changing the plan depends entirely on whether you can still change the staffing, the schedule or the flex levers in response. The combination gives a simple rule of thumb: re-forecast when the variance is genuine bias and there is still lead time to do something about it. Absent either, hold.

Tell bias from scatter before you touch the plan Noise → hold scatter both sides of plan Bias → re-forecast consistent, widening lean Re-forecast when it’s real bias and there’s still time to act.
Random scatter around the forecast is the forecast working. A persistent, widening lean is the forecast breaking — only the second is a reason to re-plan.

Make it a decision, not a reflex

Set the rule before the pressure hits: the size and persistence of variance that triggers a re-forecast, who owns the call, and the lead times of the levers you’d pull in response — so you know how late is too late to bother. Distinguish the two horizons, too: an intraday re-forecast feeds today’s flex levers and can be made on lighter evidence because it’s cheap to act on; a re-forecast of the weeks ahead changes hiring and rosters and demands a higher bar of proof. Above all, separate updating the forecast from blaming it — a forecast that’s revised on genuine new information is doing its job, not failing. The discipline isn’t forecasting harder; it’s knowing the difference between a number that’s wrong and a number that’s broken, and only moving on the second.

Pair this with forecast accuracy metrics, avoiding forecast complacency, and the discipline of deliberate inaction.