Baseline Demand

A baseline is built from periods in which the item sold at its regular price and was available, and then carried forward with trend, seasonality, price level and distribution changes taken into account. Periods distorted by the event itself are excluded – the promotional window, the forward buying that precedes it and the dip that follows it. Where an item or a store has no usable history, the baseline is borrowed from a profile of comparable stores or from a similar category. What remains is the counterfactual against which actual sales are compared.

• Uplift = actual sales − baseline
• Relative uplift = (actual sales − baseline) / baseline
• True incremental lift = uplift adjusted for cannibalization of substitutes, halo on complements and the post‑promotion dip

Baseline demand in practice

Because the baseline is a model output rather than a measurement, the method decides the verdict. Two teams evaluating the same campaign with different baseline logic report different uplifts, and neither number can be checked against reality – the world in which the promotion did not run has no data. The consequence is procedural: fix the baseline method before the campaign starts and keep it stable across evaluations, otherwise the promotion calendar is scored with a moving ruler.

Clean observations are the scarce input. Where leaflet promotions dominate the calendar, an item can go for months without a stretch of regular‑price days that is not contaminated by forward buying or by the dip that follows the campaign; stockouts remove further days, because sales recorded while the shelf was empty are censored and read as a weak baseline. The baseline also has to exist at the level at which stock moves: a campaign total says nothing about which store needs how much and on which day.

Veritico PROMO forecasts the baseline and the promotional demand separately at store–day granularity, and where history is missing it benchmarks the nearest store profiles or takes the baseline from a comparable category. Effectiveness is then measured as true incremental lift – actual sales against the computed baseline, with cannibalization, halo effect and the post‑promotion dip included – and the resulting volume feeds replenishment. Kofola raised promotional forecast accuracy to 76 % and cut the depreciation of expired stock by 14 % with Veritico PROMO (Kofola case study).

Related terms

• Promotion Planning
• Demand Forecasting
• New Product Forecasting
• Seasonality
• Time Series Forecasting
• Promotional Forecasting