Separate promotional uplift from baseline demand in distribution planning

The promotion ends, but the replenishment signal stays high

A distributor adds an uplift to the forecast for a two-week promotion. Sales increase during the campaign. The next planning run retains the uplift as if it were a permanent change, and depots continue receiving stock after the event has ended.

The problem is not simply an inaccurate forecast. It is the loss of context around the adjustment. The planning record should distinguish the underlying baseline, the temporary event assumption and the evidence used to revise either one.

This separation lets the team review whether the promotion behaved as expected without automatically carrying its peak into ordinary weeks. It also makes an extended or cancelled promotion easier to handle deliberately.

Define the baseline and the event contribution

The baseline is the expected demand without the specific promotional adjustment, under the assumptions used by the planning team. It is an estimate, not an observable certainty. The event contribution represents the additional effect the planner expects during defined periods.

The Hyndman and Athanasopoulos — Forecasting: Principles and Practice, time series regression discussion distinguishes temporary interventions from persistent changes and notes that promotional effects can extend beyond an immediate period. That provides a modelling reference, not proof that a particular campaign caused the observed sales. The planner must test the relevant pattern using appropriate data and judgement.

Record the event identifier, products, depots or channels, start and end periods, uplift assumption and owner. Include the evidence basis and any expected carryover. A free-text note attached to a total forecast is difficult to remove correctly when the promotion changes.

Keep these components visible at the level where replenishment decisions are made. A national total can hide a promotion that applies only to selected depots.

Illustrative four-week example

Assume a fictional product has a baseline of 100 units per week. A planned promotion adds 60 units in each of weeks one and two. No carryover is assumed in this simplified scenario.

For this example, assume the recorded demand is not constrained by stockouts or other fulfilment limits. The campaign forecast totals 320 units and the first two weeks’ demand also totals 320. The aggregate match hides weekly errors of ten units in opposite directions.

If the 60-unit uplift is incorrectly retained in weeks three and four, those weeks are forecast at 320 units in total against recorded demand of 200. The resulting overforecast is 120 units, or an average of 60 units per week. That is a timing and override-expiry problem even though the campaign total initially looked accurate.

The table does not prove that all demand above 100 was caused by the promotion. Other factors may have changed. It is an illustration of preserving forecast components and reviewing their periods.

Interpret sales data before learning from it

If a promoted product stocks out, recorded sales can understate demand. A low sales result should not automatically teach the next forecast that the promotion was weak. The review needs the availability history and any credible information about unfilled demand.

Likewise, a change in the number of serving depots, customer coverage or product availability can alter sales independently of the campaign. Keep those changes distinct where possible rather than assigning every movement to one promotional coefficient.

Returns, cancellations and delayed order entry can affect period comparisons. Define whether the measure is orders, shipments, net sales or another demand signal. Use the same basis for the forecast and evaluation, or explain the reconciliation.

Avoid using future information in a historical replay. The forecast version should contain only the event plan and data available at the planning cut-off. Later actuals belong in evaluation, not retrospectively in the original prediction.

Compare a dated override with a promotional model

A controlled manual uplift can be appropriate for a small number of events with clear commercial owners. Its essential controls are scope, effective periods and expiry. The weakness is maintaining those details across many products and locations.

A demand-planning platform may support scenarios and version review. The Microsoft Dynamics 365 — Demand planning overview describes such review and adjustment capabilities. Evaluate whether the actual implementation preserves separate baseline and event components and can remove a cancelled event without disturbing unrelated adjustments.

A statistical promotional model can be useful when sufficient relevant history and explanatory data exist. Its output still needs review for changed campaign design, availability constraints and uncertain future inputs. A fitted relationship is not automatic evidence of causation or guaranteed future uplift.

Compare the approaches using the same four-week example and a cancelled second week. The system should show the changed event contribution and its effect on replenishment, not simply overwrite the total forecast.

Assign expiry and review ownership

The commercial owner confirms the campaign scope and dates. The demand planner maintains the forecast assumptions and evaluates the result. Supply and depot teams assess whether the revised quantities can be executed.

An extension request should be explicit. If the promotion continues for another week, record a revised event version and the reason. Do not extend it indirectly by copying the previous week’s final forecast.

After the campaign, review the forecast at the product-location-period level before deciding whether the baseline should change. A persistent shift may justify a new baseline assumption, but it needs evidence and a documented decision.

The process should also identify leftover stock and future replenishment exposure. Removing an uplift from the forecast does not automatically cancel purchase or transfer orders already approved.

Test expiry, cancellation and recovery

Replay an event reaching its planned end. The next forecast should exclude the expired contribution while retaining the baseline and other valid adjustments. Then test an event cancelled after replenishment orders have been placed.

The planner should see those existing commitments and route any cancellation or redeployment decision to the appropriate owner. Recalculating demand alone does not reverse execution.

If event data is unavailable, show which assumptions are uncertain. Retain the last approved version for review without presenting it as current by default. On recovery, compare changes against existing supply commitments before generating new orders.

For a focused assessment, bring a redacted promotion calendar, forecast versions, availability history and resulting replenishment decisions. The review should identify whether the main weakness lies in modelling, event ownership or execution handoffs.

Promotional demand is easier to manage when the temporary assumption remains visible. That allows the business to learn from the event without letting a short campaign quietly become the permanent plan.

Next actions

1. Separate baseline, event uplift and unrelated adjustments in the forecast record.
2. Give each promotion an owner, scope and explicit effective periods.
3. Review weekly errors, availability constraints and post-event demand.
4. Test expiry and cancellation against supply orders already committed.