Promotion Forecasting: Estimating Lift and Cannibalization

Promotion forecasting, a key part of demand planning and inventory optimization, is the work of estimating how much extra volume a price cut, display, feature or bundle will generate, how much of that volume is simply borrowed from other weeks or other products, and whether the promotion makes money. It matters because promotions are one of the largest sources of forecast error in consumer businesses, and because promotions that look successful on sales alone often lose money once the discount, the dip afterward and the cannibalization are counted. A good promotion forecast separates four things: the baseline that would have sold anyway, the lift during the event, the dip after it and the effect on other products. This guide explains each, shows a worked example with checked arithmetic, including the break-even lift for a given discount, and describes how to build a library of promotion effects and use it. It follows seasonal demand planning, and leads to demand planning for CPG companies and beauty and personal care brands. Related topics include price elasticity of demand and cash flow management.

The four components

Baseline. The expected sales without the promotion, including trend and seasonality. If the baseline is wrong, every other number is wrong. Estimate it from non-promoted weeks, adjusting for season, and avoid including past promotion weeks in the baseline.

Lift. The extra volume during the promotion, often expressed as a percentage above the baseline or as a multiple. Lift depends on the discount depth, the type of promotion (price only, display, feature, multi-buy), the product, the retailer, the timing and the competition.

Dip. The reduction in sales after the promotion, as customers who stocked up buy less and others waited for the deal. The dip is larger for storable products and smaller for perishable ones.

Cannibalization and halo. The effect on other products. Cannibalization takes sales from the brand's other items or sizes. Halo adds sales of complementary items. The net effect on the brand or category is what matters to the business.

A worked example

An item has a baseline of 1,000 units a week at a price of $5.00 and a unit cost of $3.00, so the margin is $2.00 per unit and baseline weekly margin is 1,000 × 2 = $2,000 (illustrative). The promotion cuts the price by 20 percent to $4.00, leaving a margin of $1.00. During the promotion week, sales reach 2,200 units, a lift of 120 percent over the baseline (1,200 / 1,000). In each of the two weeks after, sales dip to 800 units, 200 below baseline. A sibling product with a baseline of 800 units sells 680 during the promotion week, a loss of 120 units, with a margin of $2.00 per unit.

• Promotion week margin: 2,200 × 1.00 = $2,200, against the baseline of $2,000, so +$200.
• Dip: 2 weeks × 200 units × $2.00 = -$800.
• Cannibalization: 120 units × $2.00 = -$240.
• Net effect on margin: 200 - 800 - 240 = -$840.

Check: 2,200 - 2,000 = 200, 2 × 200 × 2 = 800, 120 × 2 = 240, and 200 - 800 - 240 = -840. Although the headline lift of 120 percent looks impressive, the promotion lost $840 in margin before any trade fees or advertising costs. Sales rose, and profit fell. If the promotion had no dip and no cannibalization, it would have made $200. The sensitivity is clear: the profit from the promotion depends heavily on what happens afterward. For storable goods with a large dip, the break-even is harder to reach. For perishable or impulse items with little dip, it is easier.

The break-even lift

A discount reduces the margin on every unit, so the volume must rise enough to compensate. The break-even lift, ignoring the dip and cannibalization, is the baseline margin divided by the promotional margin. In the example, the baseline margin per unit is $2.00 and the promotional margin per unit is $1.00, so the volume must double to break even: 2.00 / 1.00 = 2.0 times the baseline, a lift of 100 percent. At a lift of 120 percent, volume is 2.2 times the baseline, so the week itself makes slightly more than the baseline margin, as the calculation above shows. A deeper discount raises the bar sharply. At 30 percent off ($3.50), the margin is $0.50, and the volume must rise to 2.00 / 0.50 = 4 times the baseline just to break even. Check: 5.00 × 0.70 = 3.50, and 3.50 - 3.00 = 0.50. Before planning any promotion, calculate the break-even lift and compare it with the lift you expect from past events. If the expected lift is below the break-even, the promotion must be justified on other grounds, such as trial, distribution, defending share or a contractual commitment.

Estimating lift from history

Build a table of past promotions with the item, retailer, dates, mechanism, depth, baseline, actual sales during the event, sales in the weeks before and after, and notes on competing events. Calculate lift and dip for each. Group by mechanism and depth, and calculate averages and ranges. For items with few events, use the group average. For new retailers or mechanisms, use judgment and start with a wide range. Statistical models, such as regression with promotion variables, can estimate effects systematically when enough data is available. Machine learning methods can handle many variables but need care and enough history. See machine learning for forecasting. Always test the model on past events it has not seen.

Cannibalization and the category view

To measure cannibalization, compare sales of related items during the promotion with their baselines. Define the groups of related items, such as sizes of the same product or competing flavors. A brand promotion that takes sales from competitors increases the brand's share, and one that takes sales from the brand's other items does not. At the retailer level, the retailer cares about category sales and margin, not the brand's alone. Understanding the retailer's view helps you design promotions that the retailer is willing to support.

Planning supply for a promotion

Promotions need supply to arrive before the event. Retailers often order ahead to build stock, so sell-in precedes sell-out. The forecast should translate the sell-out forecast into a sell-in forecast by retailer, with timing. Work out when orders will be placed and how large they will be, including the retailer's stock build. Make sure production, packaging and logistics can support the peak and that stock is available in the right locations. Plan for the dip, so that the supply does not continue at the promoted rate after the event. See demand planning for CPG companies.

Governance: measure every event

After each promotion, compare the actual with the forecast for baseline, lift, dip and net margin. Record the reasons for any difference: execution, weather, competitor action, stock availability. Use the results to update the library. Share the results with sales, marketing and finance. Over time, discussions should shift from the headline sales to the net profit of the event, and poor-performing mechanisms should be retired. Include promotions in the financial reconciliation of the plan so that volume and margin match. See financial reconciliation in S&OP.

Testing lift with stores or regions

When history is thin, a controlled test gives a cleaner estimate of lift. Run the promotion in a set of test stores and compare with similar control stores that do not run it. Suppose ten test stores average 120 units a week before the event and 300 units during it, while ten control stores average 100 before and 105 during, reflecting a mild trend (illustrative). The test ratio is 300 / 120 = 2.5, and the control ratio is 105 / 100 = 1.05. The lift attributable to the promotion is 2.5 / 1.05 = 2.38 times baseline, or a lift of about 138 percent. Check: 2.5 / 1.05 = 2.381. Using the raw test change of 150 percent would overstate the lift, because part of the increase would have occurred anyway. Match the stores carefully on size, region and past sales, and keep the test long enough to see the dip afterwards.

Fields to record for every event

A promotion library is only as good as its records. For each event, record the item and pack, retailer or channel, start and end dates, mechanism, depth, price, features such as display or flyer, competitor activity if known, baseline estimate, sales during, sales before and after, stock availability, and the margin result. Add a note on anything unusual, such as weather or stockouts. Store the records in a table that planners and finance can both read. Consistent fields make it possible to compare events, and a few hundred events are usually enough to see which mechanisms work for which items.

Linking the promotion calendar to the plan

The promotion calendar should be an input to the consensus plan, not something that sales and marketing manage separately. Review it monthly with a view of the next three to six months. Check each planned event for the break-even lift, supply readiness and overlap with other events, since several promotions in the same category in the same weeks reduce each other's effect. Where events are planned at the retailer's initiative, include the retailer's expected volume and timing. Remove or modify events that are unlikely to pay back, and record the decision.

Common mistakes

Measuring a promotion by sales during the event alone ignores the dip and cannibalization. Including past promotion weeks in the baseline overstates the baseline and understates lift. Using average lift for all retailers hides differences. Ignoring stock availability means lift is capped by supply. Failing to calculate break-even lets unprofitable promotions continue. Not measuring after the event prevents learning.

FAQs

What is promotion lift?

The extra volume during a promotion compared with the baseline, usually shown as a percentage or a multiple.

How do you calculate the break-even lift?

Divide the baseline margin per unit by the promotional margin per unit. The result is the multiple of baseline volume needed to break even, ignoring dip and cannibalization.

What is the post-promotion dip?

The reduction in sales after an event as customers who stocked up buy less or others delayed purchases.

How do I measure cannibalization?

Compare the sales of related items during the promotion with their baselines, and sum the losses.

How often should promotion effects be updated?

After every event, and the library should be reviewed at least once a quarter.