Skip to content
Predict.ai

Decisions

Capacity planning: the constraint is the forecast

A plant, a ward, a classroom, a network link — capacity is a ceiling. The forecast is whether you will hit it, and when. Over-build is capital. Under-build is a queue, a blackout, or a denied admission.

Updated Aug 26, 2026·7 min read

Forecast the load, know the ceiling

A plant, a ward, a classroom, a network link — capacity is a ceiling. The forecast is whether you will hit it, and when. Over-build is capital. Under-build is a queue, a blackout, or a denied admission. Forecast the work (orders, patients, packets, enrollments). Utilization is a ratio that moves when either side changes. You will not know which.

Two horizons

Adding a line, a wing, or a fiber build has a long lead. Running this week’s shift has a short one. Those are different goals. Stretching the daily production path to 36 months is how you get a confident expansion case on a model that never saw a cycle.

  • Short: can we clear this week’s load with the lines we have?
  • Long: a wide interval, plus scenarios for demand and for a line down.
  • Queues hide the miss. If you only see wait time, you are late.

Queues hide the miss

Hospitals divert. Networks drop. Factories build WIP. The demand still happened. Training on the throttled output teaches the model your constraint, not the load. Flag the throttle, or forecast arrivals / orders, not throughput, when the ceiling binds.

Base path vs promo-on

What-if scenario versus base forecastdecisionpromo onbase
Same model, same horizon. Lime is the scenario ('run the 10% discount from Friday'). The gap is the estimated lift — still a band in production, drawn here as two center lines so you can see the fork.

Load as a goal, capacity as a scenario

Forecast the load in Predict.ai. Then run a scenario with a line down, a classroom closed, or a link degraded. You see whether the path still clears the constraint — before you buy a tool or deny a cohort. Alerts when load leaves the band are how you notice the squeeze without waiting for the queue to become a scandal.

FAQ

Is utilization a good target to forecast?
Usually no. Forecast the work (orders, patients, packets). Utilization is a ratio that moves when either the numerator or the denominator changes. You will not know which.
How do I plan a factory expansion?
A long-horizon, wide-interval forecast of output or demand, plus a scenario for the new line. Do not use the daily production path stretched to 36 months.

Keep going

Try it on your data.

Connect an outcome and its history. Predict.ai finds the drivers, races the models, and keeps the forecast live — in the workspace, over the API, and through agents.