Inventory and Planning

We Plan to a Forecast That Is Always Wrong

Every forecast is wrong; the question is by how much and in which direction. A plan that treats the forecast as exact pushes the error onto the floor and the suppliers as expedites and overtime. Measuring the error and planning around it works better than forecasting harder.

What Is Usually Going On

  • Forecast error is never measured, so nobody knows how wrong is normal.
  • Lead times are longer than the window the forecast is good for.
  • The plan is built to the single forecast number with no buffer for its error.
  • The plan changes inside the frozen window, so every change hits the floor.
  • Demand, supply and capacity are planned in separate meetings.

How to Fix It

  1. Measure the errorForecast against actual by item family each period. Track the average error and the bias (whether it runs high or low).
  2. Fix bias firstA forecast that is always high or always low is a process problem, not noise. Correct it before adding buffer.
  3. Buffer by the errorSize safety stock from forecast error over the lead time, not from demand history alone.
  4. Shorten the lead timeThe shorter the lead time, the less forecast you need. Cut queues and batch sizes so the plan can follow real orders.
  5. Freeze the near termSet a frozen window the floor and suppliers can execute. Changes inside it need one named approver.
  6. Plan to demonstrated capacityLoad the plan against what the area has actually shipped, not rated capacity. Settle the gaps in one monthly planning meeting (S&OP).

Start Free

Guides and web tools on this site. No account needed.

Guide
S&OP: Sales & Operations Planning
Guide
Inventory Management for Manufacturing
Guide
Push vs. Pull Production Systems
Calculator
Safety Stock Calculator
Calculator
WIP and Little's Law Calculator

When It Moves to the Floor

The same work as an offline file you keep: saved scenarios, print and export, no internet needed on the floor.

Common Questions

What is a good forecast error?

There is no general number. Compare your error by family over time, and against a naive forecast that uses last period's actual. If you cannot beat the naive one, simplify.

Should we forecast in more detail?

Usually not. Forecasts hold up better at family level and over longer buckets. Forecast the family, then split by mix closer to the date.

What is S&OP?

Sales and operations planning: one monthly meeting where demand, supply and capacity are matched and the gaps decided, instead of each function planning alone.

Can pull replace the forecast?

For repeat items with short lead times, pulling from actual use cuts how much you depend on it. Long-lead items still need a forecast.