Historical Data Based Forecasting

Historical Data Based Forecasting is the approach of predicting future delivery outcomes by directly extrapolating from a team's own recorded past performance, using metrics such as velocity, throughput, or cycle time gathered over prior iterations as the primary evidence base for projecting what is likely to happen going forward. It stands in contrast to forecasting methods built on manual, item-by-item estimation, instead grounding predictions in what the team has actually and demonstrably done, consistent with the broader purpose of forecasting introduced earlier in this topic area.

The Underlying Premise

Past Performance as the Best Available Predictor

The central assumption behind this approach is that a team's recent historical performance, assuming its process and context remain reasonably stable, is a more reliable guide to future performance than an estimate constructed independently of that history, since actual outcomes reflect the full complexity of the team's real working conditions in a way that a hypothetical estimate cannot fully anticipate.

Reducing Reliance on Speculative Estimation

Rather than requiring the team to predict in advance how long each individual future item will take, historical data based forecasting treats the team's own aggregate track record as the source of truth, sidestepping much of the speculative judgment involved in item-by-item estimation and instead relying on demonstrated, measurable results.

Common Approaches Within Historical Data Based Forecasting

Simple Average Projection

The most basic form of this approach projects future delivery by applying the team's average historical velocity or throughput directly, dividing the remaining scope of work by that average rate to estimate a completion timeframe, a straightforward calculation that nonetheless carries the limitation of representing only a single point estimate rather than a range.

Probabilistic Simulation

A more sophisticated approach repeatedly samples from the team's actual historical delivery data to simulate many possible future outcomes, producing a range of plausible completion dates along with the relative likelihood of each, offering a fuller picture of uncertainty than a single average-based projection can provide, and this technique is developed further as a dedicated forecasting method elsewhere in this practice area.

Trend-Adjusted Projection

Where a team's recent performance shows a clear directional trend, such as steadily increasing throughput following a recent process improvement, a trend-adjusted projection weights more recent periods more heavily than older ones, aiming to reflect the team's current trajectory rather than treating its entire history as equally representative of what to expect going forward.

Selecting the Historical Window

Balancing Recency Against Sample Size

Choosing how much historical data to include in a forecast involves a trade-off: a longer window provides a larger, more statistically stable sample, while a shorter window better reflects the team's current process and conditions if those have changed meaningfully since the earlier data was recorded.

Excluding Data From a Different Process Era

Historical data gathered before a significant, deliberate process change, such as a team composition shift or a major workflow redesign, generally should not be blended with more recent data as though the two periods represented the same underlying process, since doing so would combine two genuinely different patterns into a single, less meaningful average.

A Historical Throughput Sample Used for Forecasting

The bars within the highlighted window show a noticeably higher and more stable throughput than the earlier bars outside it, illustrating why the choice of historical window can materially affect a resulting forecast if the excluded earlier data reflects a meaningfully different period.

Validating Forecast Reliability Over Time

Comparing Past Forecasts Against Actual Outcomes

A historical data based forecasting practice becomes more trustworthy over time when the team periodically checks its previous forecasts against what actually occurred, a discipline directly connected to the forecast accuracy tracking introduced in Delivery Predictability Assessment.

Adjusting the Approach Based on Observed Accuracy

Where forecasts have consistently proven overly optimistic or overly conservative relative to actual outcomes, the team adjusts its chosen historical window, weighting scheme, or underlying method accordingly, treating the forecasting approach itself as subject to the same kind of iterative refinement applied to any other team process.

Common Pitfalls

Treating a Single Point Estimate as Sufficiently Informative

Relying solely on a simple average projection, without any accompanying sense of the underlying variability, presents an illusion of precision that the team's actual historical data, often quite variable, does not genuinely support.

Blending Data Across a Significant Process Change

Combining historical performance from before and after a major process shift into a single averaged forecast basis produces a projection that does not accurately represent either period, weakening its practical usefulness.

Ignoring Signals That the Historical Pattern No Longer Applies

Continuing to base forecasts on a historical dataset without checking whether recent actual performance still resembles that history risks producing increasingly inaccurate projections as the team's real circumstances diverge from the conditions under which the historical data was originally recorded.