Short answer: Promotional volume spikes are real but routinely misinterpreted. The discipline of promotional lift forecasting separates genuine demand creation from timing effects, forward-buying, and cannibalization of adjacent SKUs. Without this separation, promotional ROI is systematically overstated and supply plans built on promoted volumes will create costly overstock in the weeks that follow.
Why Promotions Break Naive Forecasts
A promotion creates a pattern in shipment and depletion data that looks, superficially, like demand growth. Units sold spike during the promoted week or weeks, then fall back — sometimes below the pre-promotion baseline as retailers and consumers draw down the stocks they accumulated during the deal.
A naive forecasting model trained on this raw signal learns the wrong lesson: it expects a permanent demand increase that is not there. The result is over-production commitments, elevated finished goods inventory, and write-downs.
The fix is not to ignore promotions — it is to model them explicitly as a separate demand component rather than absorbing them into the baseline trend.
The Three-Component Framework
A robust promotional lift model decomposes total sold volume into three components:
1. Baseline demand. The volume that would have been sold at full price, absent any promotional activity. This is the number that reflects genuine consumer preference and is the right input for capacity planning, raw material procurement, and long-range financial forecasts.
2. Incremental lift. The volume genuinely added by the promotion — either from consumers buying the category more frequently (category expansion) or switching from a competitor (trade-up). This is the value the retailer and brand are trying to create.
3. Timing transfer. Volume shifted from adjacent periods — forward-buying by retailers filling displays, consumers stocking up, or pantry-loading. This volume will be negative in the post-promotion period. It creates no net demand.
The practical challenge is that these three components are not directly observable; they must be estimated. Regression-based models that control for promotional flags, price indices, and retailer inventory estimates can decompose the signal with reasonable accuracy for well-established SKUs with long histories.
Measuring the Post-Promotion Dip
The most under-monitored metric in promotional analytics is the post-promotion trough. A credible lift analysis always shows the two-to-four week period following a major promotion alongside the promoted period itself.
The ratio of the dip to the lift indicates how much of the promotional volume was borrowed from the future versus genuinely incremental. A promotion that generates a 25% volume uplift in week one but a 15% dip in weeks two and three has delivered much less net value than it appears at first glance.
For supply planning, this means production schedules should not respond symmetrically to promotions. Producing to the promoted peak and then cutting aggressively post-promotion is typically the right approach — but only if the planning system is receiving accurate promotional calendars in advance rather than reacting to shipment spikes after the fact.
Cannibalization Within the Portfolio
Promotions on one SKU can suppress sales of adjacent SKUs in the same portfolio. A heavy discount on a flagship lager may slow sales of a premium craft variant for the same brand. For a brewery managing both conventional and non-alcoholic SKUs, this within-portfolio cannibalization is particularly important to measure.
See Cannibalization: Does Non-Alcoholic Beer Eat Your Lager’s Sales? for a framework specific to the NA-versus-conventional dynamic.
The principle applies equally to promotional context: if a promotion on the conventional lager accelerates the category trial that later converts to NA purchasing, that is a positive spillover. If it simply delays an NA purchase by a few weeks, it is a timing transfer. Distinguishing the two requires either panel data or a carefully designed test-and-control market structure.
Forecasting Forward: Building the Promotional Calendar Into the Model
A promotional lift forecast is only useful if the forward promotional plan is available as an input. This sounds obvious and is routinely violated. Promotions are frequently negotiated and confirmed by sales teams on timelines that do not align with supply planning lead times.
The organisational fix is a locked promotional calendar — ideally frozen four to eight weeks ahead — that feeds directly into the demand model. Any promotion confirmed after the lock-in date should trigger an explicit discussion between sales and supply about whether volume commitments can be adjusted within available capacity and inventory.
For a broader view of how promotional modelling fits within the driver-based forecasting stage, see The Beverage Demand Forecasting Maturity Model.
The Honest Caveat
Promotional lift models require enough promotional events in the historical dataset to estimate lift coefficients reliably. For SKUs that are promoted infrequently, or for NA beer SKUs with limited history, the model will have wide uncertainty intervals around any lift estimate. In those cases, a simpler rule-of-thumb (e.g. assume X% lift based on category analogues) may be more honest than a fitted model that appears precise but is extrapolating from thin data.
Part of the Sales Forecasting track — browse all.
Frequently asked questions
What is promotional lift in demand forecasting?
Promotional lift is the incremental volume sold during a promotion above what would have been sold at the regular price and conditions — the baseline demand. Isolating it is essential because promotions create volume spikes that, if mistaken for genuine demand growth, will cause over-production and inventory problems after the promotion ends.
How do breweries typically over-estimate the value of a promotion?
The most common error is comparing the promoted period to the prior period without accounting for forward-buying and post-promotion dips. Consumers and retailers stock up during a deal, pulling volume forward; the weeks immediately after a promotion often show below-normal sales. Looking only at the lift without netting the dip significantly overstates the net demand created.
Do promotions work differently for non-alcoholic beer?
The evidence is directionally mixed. NA beer tends to attract trial-oriented buyers who may be more price-sensitive at the point of first purchase, suggesting promotions can drive genuine new-buyer acquisition rather than just forward-buying. However, the category is still maturing and the data to make confident generalisations is limited for most brands.