Uncertainty and models

Fan chart (forecast uncertainty)

Displays forecasts with nested probability or quantile bands around possible outcomes.

  • Chart
  • Uncertainty
  • Forecasting

What it answers

What forecast outcomes are plausible and how uncertain are they?

Shape
Line with nested forecast bands
What it shows
Central trajectory and changing forecast spread.

Example

The question

How wide is the range of possible revenue next year?

Quarterly revenue with forecast range (EUR million)

Quarterly revenue with forecast range (EUR million)Actual revenue rises from 10 to 13.6 by Q4 2025. The central forecast reaches 15.3 in Q4 2026; the inner band (likely range) runs 13.6 to 17.0 and the outer band 11.9 to 18.7.ActualForecastEUR million6101418Q1 24Q2Q3Q4Q1 25Q2Q3Q4Q1 26Q2Q3Q4
Quarterly revenue with forecast range (EUR million)Actual revenue rises from 10 to 13.6 by Q4 2025. The central forecast reaches 15.3 in Q4 2026; the inner band (likely range) runs 13.6 to 17.0 and the outer band 11.9 to 18.7.ActualForecastEUR million6101418Q1 24Q3Q1 25Q3Q1 26Q3
Actual revenue rises from 10 to 13.6 by Q4 2025. The central forecast reaches 15.3 in Q4 2026; the inner band (likely range) runs 13.6 to 17.0 and the outer band 11.9 to 18.7.

Illustrative data. It belongs to no company.

When to use it

Use it when

Communicating forecast distributions over a horizon.

Avoid it when

When bands lack coverage definitions or are mistaken for guarantees.

How to read it

Read the central forecast and the probability labels on nested bands.

Note

State interval probabilities and what they cover.

The data you need

Layout
Wide interval table
Fields
Time: date; Forecast: number; Lower/Upper columns: paired numbers for each band
One row is
One forecast period with its central estimate and interval bounds. What is the grain?
Rules
Pair and label each interval coverage. Require Lower95 <= Lower80 <= Upper80 <= Upper95 for nested central bands.
Optional
Historical Actual values, model/version and forecast origin. Intervals must come from an appropriate model.
Preparation
Fit or obtain a suitable model/estimate first, then pair estimates with correctly labelled bounds or standard errors. Preserve coverage and scale information.
Units
Forecast value and bounds use native outcome units. Band percentages specify model coverage, not percentage of observed records.
Example of the input layout: Wide interval tableA wide table: one row per item, with a separate column for each measure or period.TimeV1V2V3V4V5ABC
A wide table: one row per item, with a separate column for each measure or period.

Worked example

Forecast 100 with 80% band 90-110 and 95% band 80-120. Draw nested bands. The wider band has greater coverage under the model.

TimeForecastLower80Upper80Lower95Upper95
2026-11-011009011080120
2026-12-011058812272138