Skip to content

MODULE 10 · LESSON 58 · Advanced application

Forecasting limitations

Distinguish a projection based on assumptions from an observed fact and evaluate forecasts against a baseline.

By Laurence Svekis · Self-paced · All worked data is synthetic unless you supply your own source.

Try the practice ↓

PredictInspectCalculateCheckExplainTransfer

Your starting point

Prerequisites: Use the ideas and vocabulary from lesson 57; revisit its worked examples if needed.

Definition: A forecast estimates a future outcome using a stated method and available information. Extrapolation extends a pattern beyond the observed range. A scenario explores what could happen under assumptions; it is not necessarily a probability-based prediction.

State the rule and its assumptions

A straight-line continuation assumes that the recent pattern remains useful. That can fail when seasonality, capacity, prices, definitions or behavior change. A model can fit historical observations closely without predicting new ones well. Keep the forecasting rule simple enough to explain and compare it with a basic baseline such as the most recent observed value.

Evaluate on information not used to fit

When possible, use an earlier period to form the prediction and a later observed period to check it. Avoid using future information accidentally through revised records or calculations that include the outcome you are predicting. One successful prediction is weak evidence; inspect errors across several relevant periods and contexts. The academy does not offer a black-box forecasting service.

Communicate uncertainty and decision boundaries

Present assumptions, plausible alternative scenarios and conditions under which the projection should be revised. Do not attach invented confidence intervals to a deterministic rule. A forecast for planning can be useful even when uncertain, provided the action remains proportionate and the model is reviewed as new evidence arrives. Never describe projected values as if they were already measured observations.

Two worked examples

A simple continuation

Observed monthly totals are 100, 110 and 120. Continuing the ten-unit increment gives a next-period projection of 130. A last-value baseline gives 120. Neither is a fact about the future. A holiday, capacity limit or measurement change could invalidate the continuation assumption.

A holdout check

Use January through April to predict May, then compare the prediction with May’s observed value. If you had already used May to choose the rule, that would not be an independent check. Record the information available at prediction time and compare the error with a simple baseline.

Where this becomes useful

A small organization plans staffing using recent demand. A range of explicitly assumed scenarios may support a reversible plan better than a single precise number presented with unjustified confidence.

Make an attempt

Use the Statistics Calculator or workbench to create a short synthetic time series. Write a last-value baseline and a simple trend projection before revealing a held-out value. Compare errors and explain one condition that could break the trend.

Open the accompanying interactive lab →

Misconception check

Claim: A smooth historical line guarantees a reliable forecast.

Inspect the correction

Future conditions may differ, and fitting past observations is not the same as predicting unseen outcomes.

Independent application

Build a small synthetic forecasting exercise with an explicit rule, a last-value baseline and a held-out observation. Calculate both errors, describe the assumptions and write a planning recommendation that can be revised when new evidence arrives.

Completion records practice and self-review. It is not certification of statistical competence.