Promotional Forecasting

The result is not one number but a set of them: a total for the campaign, a split per store and a profile per day, because the time profile of the uplift depends on the mechanics – where the leaflet lands at the start of the week, the profile is front‑loaded. Alongside the uplift itself, the forecast has to account for effects that sit outside the promoted item: cannibalization of substitutes in the same category, halo on complementary products, and the dip that follows the campaign when customers who bought ahead stop buying.

• Drivers: discount depth, mechanic (multibuy, buy & get), leaflet position (cover versus inside page), secondary display, TV or online support.
• Granularity: store–day, because replenishment consumes what has to arrive on which shelf on which day, not the campaign total.
• Side effects: cannibalization within the category, halo effect, post‑promotion dip.
• Baseline dependency: the uplift is only as good as the baseline it is measured against.

Promotional forecasting in practice

The forecast is produced while its inputs are still decisions. A leaflet closes weeks before the campaign, and between the close and the first day the discount depth, the page position or the media support can still change; each change moves the uplift, and a forecast frozen at leaflet close describes a promotion that may no longer exist. A workable setup therefore treats the forecast as a function of the mechanics, re‑run whenever they change, rather than as a number attached to the campaign.

The second issue is the level at which the forecast is judged. A campaign total that comes out right can still leave part of the network without stock from day three and the rest with leftovers to mark down, because the split across stores and days was wrong; accuracy measured on the total does not see this. The error also lands on a larger volume than a baseline error: where the promoted volume is three times the baseline, the same relative error means three times as many units short or left over. That is why promotional forecast accuracy is reported separately from baseline accuracy, and why a value that looks modest next to a baseline figure can be a good result.

Veritico PROMO forecasts baseline and promotional demand separately at store–day granularity; where an item or store has no comparable history, it benchmarks the nearest store profiles or takes the baseline from a similar category, and the campaign volume is handed to replenishment. Effectiveness is then read as true incremental lift against the computed baseline, including cannibalization, halo effect and the post‑promotion dip, and the model quantifies the effect of placement and media – TV, leaflet cover – on the uplift; see promotion planning and effectiveness. Kofola reached 76 % promotional forecast accuracy with Veritico PROMO and cut the depreciation of expired stock by 14 % (Kofola case study). Where promotions are not flagged in the data, Veritico STOCK detects them automatically and cleans the history before forecasting, so the baseline is not inflated by unlabelled campaigns – see demand forecasting and inventory optimization.

Related terms

• Promotion Planning
• Baseline Demand
• Price Elasticity of Demand
• Forecast Accuracy