A forecast-versus-actual bridge explains a difference using defined, non-overlapping components. Preserve the original forecast and identify changed assumptions, timing shifts, mix effects and corrections. Reconcile the bridge to the actual total without treating every unexplained remainder as operational performance.
Freeze the forecast being evaluated
Record the forecast version, creation date, source cut-off and assumptions. A revised forecast can help current planning, but it should not replace the original when evaluating what was expected earlier. Define the unit and horizon consistently with the actual result. Forecast bookings cannot be compared with collected cash without an explicit bridge between the events.
Choose the actual-data version and note whether it is provisional. Late records or corrections may change the apparent variance after the first review. Keep these changes visible as revisions rather than evidence that operations changed retroactively. If the historical forecast inputs are unavailable, state the reconstruction limit. The evaluation should distinguish what was predicted from what can now be calculated with hindsight.
Build components with a stated calculation convention
Start with drivers that have an operational interpretation, such as eligible demand, conversion, units per order or completion timing. Determine an order for changing assumptions and disclose it, because interaction terms can be allocated differently by different bridge conventions. Each component should have a calculation, evidence and unit. Do not use labels such as market impact for a residual whose cause has not been established.
Separate timing differences from permanent changes where evidence supports that distinction. Work delayed into the next period is not necessarily lost, but a proposed later date is not proof it will occur. Keep pending commitments visible and review them later. A data correction should have its own line so source repair is not mistaken for improved operational performance.
Use the bridge to improve the next decision
Prioritise components that are material and actionable. An incorrect demand assumption may require a different source; a delivery constraint may need capacity planning; a recurring late-data adjustment may need a reporting-process fix. Assign an owner and review date to each important investigation. Do not expect the bridge itself to establish causation merely because its arithmetic reconciles.
Retain an unexplained residual when evidence is insufficient. Reconciliation without explanation is still useful if labelled honestly, while forcing a convenient explanation can misdirect action. Compare later outcomes with timing assumptions and update the learning record. The final brief should show the original forecast, the actual version, the reconciled components and the uncertainty that matters for the next planning cycle.
A fictional completion forecast bridge
A fictional scenario to explain the method, not a customer result or a market benchmark.
A fictional training service forecasts 1,000 completed learner modules in October. The agreed bridge identifies 80 fewer completions from lower eligible enrolment, 50 modules deferred to November because a session moved, and 20 additional completions from a late correction to source records. Applying the signed components gives 1,000 minus 80 minus 50 plus 20, or 890 actual recorded completions.
The total shortfall is 110 modules. The team does not describe all 110 as lost demand: 50 are a documented timing deferral and 20 offset the shortfall through a data revision. It checks the enrolment calculation and monitors whether the deferred session actually occurs. If it later cancels, the explanation is revised with a date instead of silently rewriting October's forecast. The bridge uses fictional values and a stated component convention; arithmetic agreement does not independently verify the operational causes assigned to each line. The analyst retains the twenty corrected records and their availability dates so a reviewer can confirm that the correction concerns October completions. It also checks that the fifty deferred modules are not counted again among the eighty attributed to lower enrolment. This avoids a bridge that reconciles only because overlapping explanations were combined.
TAKE THIS INTO YOUR NEXT CONVERSATION
A practical next-step checklist.
- Freeze forecast version and information cut-off.
- Match forecast and actual units and horizon.
- Separate timing and data corrections.
- Retain unexplained residuals with an owner.
Common questions.
Should the latest forecast replace the original in an accuracy review?
No. Preserve the original for evaluation and show later revisions as separate planning versions.
Does a bridge that sums correctly prove its explanations?
No. Each labelled driver needs supporting evidence, and interaction allocation depends on the chosen convention.
Further reading & source context.
These primary references provide background. They are not evidence for the fictional example above and do not endorse Accuralytix.
Found an error or a missing qualification? Send an editorial correction.
Check my data readiness.
Use a practical tool to turn this idea into a clearer starting point.
Check my data readiness