Forecast Range and Confidence

Forecast Range and Confidence concerns how the width of a forecast's stated range and the confidence level attached to it should be selected and interpreted, addressing the practical judgment calls that come after a probabilistic forecast, such as one produced through Monte Carlo Forecasting, has already generated its underlying distribution. Producing a distribution of possible outcomes is only the first half of delivering a useful forecast; deciding which portion of that distribution to report, and how to explain what the chosen confidence level actually means, determines whether the forecast is genuinely useful to the people relying on it.

The Relationship Between Range Width and Confidence Level

Wider Ranges Carry Higher Confidence

A forecast range constructed to capture a higher percentage of simulated outcomes must necessarily be wider than one constructed to capture a lower percentage, since including more of the distribution's spread, particularly its less common but still possible extremes, requires extending the range further from its central tendency.

The Fundamental Trade-Off Between Precision and Reliability

A narrow range feels more useful for planning because it is more specific, but a narrow range drawn from a probabilistic distribution necessarily corresponds to a lower confidence level, meaning the actual outcome has a correspondingly higher chance of falling outside it; forecast range and confidence is fundamentally about navigating this trade-off deliberately rather than by accident.

Common Confidence Levels and Their Practical Meaning

Lower Confidence Levels for Central Tendency

A confidence level around the fiftieth percentile represents the median simulated outcome, useful as a description of a typical or central case but carrying, by definition, roughly even odds that the actual result will fall earlier or later than this point, making it a poor choice as the sole basis for an external commitment.

Higher Confidence Levels for Risk-Aware Commitments

A confidence level around the eighty-fifth or ninetieth percentile provides a more conservative reference point, capturing a substantially larger share of plausible outcomes and therefore better suited to situations where the cost of being late significantly outweighs the cost of a more cautious estimate.

Very High Confidence Levels for Critical Commitments

For commitments where missing the forecast carries especially serious consequences, teams sometimes report a confidence level near the ninety-fifth percentile or higher, accepting a considerably wider range in exchange for a stronger guarantee that the actual outcome will fall within it.

Selecting an Appropriate Confidence Level for the Situation

Matching Confidence to the Cost of Being Wrong

The appropriate confidence level to report depends on the consequence of an inaccurate forecast in the specific context at hand, with situations carrying a high cost of lateness warranting a higher, more conservative confidence level than situations where an approximate, central estimate is sufficient for planning purposes.

Considering the Audience's Sophistication

An audience familiar with probabilistic reasoning may be well served by seeing multiple confidence levels presented together, while a less technically oriented audience may need a single, carefully chosen confidence level accompanied by a clear, plain explanation of what it represents, since presenting excessive statistical detail to an unprepared audience can obscure rather than clarify the forecast's practical meaning.

Visualizing the Range and Confidence Trade-Off

Each successive range fully contains the narrower ranges within it, illustrating that higher confidence is achieved specifically by widening the reported window rather than by any improvement in the underlying forecast's accuracy.

Expressing Range Width Quantitatively

The width of a reported confidence range can be expressed directly as the interval between its lower and upper percentile bounds.

Tracking how this width changes across successive forecasts of similar work can itself serve as a secondary indicator of the team's underlying predictability, since a persistently wide range, even at a fixed confidence level, reflects the same kind of high variability discussed under Delivery Predictability Assessment.

Communicating Range and Confidence Honestly

Avoiding the Appearance of False Precision

Reporting a forecast range with an implied but unstated confidence level, or presenting only a narrow range without disclosing that it corresponds to a relatively low confidence level, misleads stakeholders about how much certainty the forecast actually carries.

Explicitly Naming Both the Range and Its Confidence Level Together

Effective communication always pairs the two pieces of information, stating both the specific date range and the percentage of simulated outcomes it represents, giving the audience everything needed to correctly judge how much weight to place on the forecast.

Common Pitfalls

Selecting a Confidence Level Without Considering Context

Defaulting to the same confidence level for every forecast regardless of the actual stakes involved fails to account for the fact that different situations warrant different trade-offs between precision and reliability.

Presenting a Narrow Range as Though It Were High Confidence

Choosing to report a narrow, appealing-looking range without disclosing that it corresponds to a modest confidence level sets stakeholders up for disappointment when the actual outcome, quite plausibly, falls outside that narrow window.

Treating the Confidence Level as a Guarantee

Even a very high confidence level, such as ninety-five percent, still carries a real, non-zero chance that the actual outcome falls outside the stated range, and communicating the range as an absolute guarantee misrepresents the fundamentally probabilistic nature of the underlying forecast.