In brief: Sales reports for the campaign week mostly present a positive picture. However, the sales decline in the following weeks is usually not reflected in this picture. Although campaigns make up a significant share of sales in fast-moving consumer goods, not all of the increase in the campaign week can be counted as additional demand; part of the increase is sales pulled forward from the post-campaign weeks. In addition, each campaign type affects demand differently: while a price discount creates a fast, short-lived jump, sampling or advertising campaigns show their effect more slowly and over a longer period. This article covers the components of the promotion effect, the effects on demand by campaign type, methods for calculating the effect, and the tools used in modeling. The general planning problems of the sector are examined separately in the article Demand Planning and S&OP in Fast-Moving Consumer Goods (FMCG): 5 Critical Topics. In this article:What Components Make Up the Promotion Effect?How Do Campaign Types Affect Demand Differently?How Can Campaign Effects Be Calculated?How Is a Campaign Forecast Built?Why Is Post-Campaign Analysis Necessary?How Is the Campaign Calendar Linked to S&OP?Technology and Digitalization: Tools Used in Modeling Campaign EffectsImplementation: Checklist What Components Make Up the Promotion Effect? The promotion effect is the entire change a campaign creates in demand. From a planning perspective, expressing this change as a single figure is not sufficient, because each component affects a different point of the plan. Baseline demand is the sales that would have occurred without the campaign. This is the layer the statistical forecasting model should produce, and it enters the plan as the foundation. Gross uplift is the sales realized above baseline demand during the campaign period; it is reflected in the production and shipment plan as a separate layer added on top of baseline demand. Gross uplift gives the first impression of the campaign's success; but on its own it is misleading. Pull-forward is the sales drawn from later weeks because consumers and the channel also buy their future needs during the campaign. Its result is the post-campaign dip, and for this reason the forecast for the following weeks must be pulled below baseline demand. The dip is pronounced for stockable products and deep discounts, and limited for perishable products. Channel stocking is the loading of distributors and the traditional channel with campaign volumes before the sale to the consumer; the time lag between sell-in and sell-out must therefore be monitored separately. When these components are read together, the true contribution of the campaign is the net incremental sales, found by subtracting pull-forward and the post-campaign dip from gross uplift. Planning decisions (production quantity, capacity, materials, stocking) rely on each of these three layers being forecast correctly and separately. How Do Campaign Types Affect Demand Differently? In practice, "campaign" is not a single type, and the effect profile of each type on demand (the intensity and duration of the uplift, the dip afterwards, and persistence) is different. Using a single campaign coefficient therefore leads to systematic deviations in the forecast. The main campaign types and their typical effects on demand are covered below. Temporary price discount: The most common type. The effect appears as a fast, high jump in the campaign week and grows as the discount depth increases. For stockable products, the post-campaign dip is pronounced. Multi-buy campaigns (buy 3 pay for 2, discount on the second item): They increase basket size, but a significant part of the increase comes not from higher consumption but from stockpiling. For this reason, the post-campaign dip lasts longer than with a single-product discount. Free gift and bundle campaigns: Since they do not include a price discount, the increase in demand is usually more moderate. The main point to watch in planning is that the additional product inside the gift or bundle is also included in the supply and inventory plan. Display, catalog, and flyer campaigns: Demand is increased through visibility without a price change. The effect depends on the rate at which the display is actually implemented at points of sale; if the implementation rate is lower than planned, the expected uplift does not materialize. Coupon, loyalty point, and cashback campaigns: The effect appears only in the customer segment that uses the coupon or points. Since the redemption rate is uncertain, the forecast range is wide and the effect may be spread over time. Advertising and digital marketing campaigns: The effect is mostly delayed and diffuse; it can continue for some time after the campaign ends. The relationship between advertising intensity and sales is not linear, and returns diminish beyond a certain level. Sampling and launch campaigns: Trial distribution may lead not to a post-campaign dip but, on the contrary, to repeat purchases by customers who tried the product. In this case, the effect is not temporary but appears as a permanent rise in baseline demand. Distributor and channel campaigns (trade terms, volume bonuses): The effect first appears in sell-in data. Since consumer sales mostly occur with a delay, channel stock swells during the campaign and orders fall in the following period. End-of-season clearance campaigns: The goal is to run down stock. Demand jumps, but margin is low and it may affect the baseline demand of the next season. For this reason, the type, mechanic, channel, and duration of each campaign should be stored as separate fields in the campaign library, and effect calculations should be made by type. The same campaign type can give different results in different categories. How Can Campaign Effects Be Calculated? The basic logic of calculating the campaign effect is simple: the effect is the difference between actual sales and the sales that would have occurred without the campaign. The difficulty is that the second value can never be observed directly. For this reason, the sales expected without the campaign (baseline demand) must be forecast separately. The subject can be explained through an example. Assume that a product with weekly baseline demand of 1,000 cases is put on a campaign for one week (figures are for illustration purposes). 2,600 cases were sold in the campaign week, and 800 and 850 cases in the first and second weeks after the campaign. Gross uplift is 2,600 − 1,000 = 1,600 cases. Since the dip in the two weeks after the campaign is (1,000 − 800) + (1,000 − 850) = 350 cases, net incremental sales are calculated as 1,600 − 350 = 1,250 cases. The post-campaign dip is about one fifth of gross uplift. In evaluating net incremental sales, not only quantity but also the financial result is taken into account. The margin on net incremental sales is compared with the cost of the campaign (discount amount, display fee, additional logistics). If the discount applies to the entire base quantity that would have sold anyway, the campaign can produce a loss when net incremental sales are not high enough. For this reason, both the financial evaluation of the campaign and the forecast of subsequent similar campaigns should be based on net incremental sales, not gross uplift. Methods Used in the Calculation Baseline subtraction method: Campaign periods are cleansed from historical series and baseline demand is estimated statistically; the effect is found as the difference between actual and this estimate. It is the easiest method to apply, but the result depends on the quality of the baseline demand estimate. Control group comparison: The sales changes of stores, customers, or regions where the campaign is applied are compared with similar units where it is not. By taking the difference of the differences before and after the campaign (the difference-in-differences method), seasonal and general effects are excluded. Where possible, it is the approach that gives the most reliable result. Regression models: Sales quantity is linked to explanatory variables such as price, discount rate, display, advertising intensity, and seasonality. The coefficients give the effect of each variable and price elasticity (by what percent demand changes when price changes by one percent). It requires a sufficient and varied campaign history. Similar campaigns (analog approach): The effect of the new campaign is derived from the average and distribution of past campaigns with the same mechanic, channel, and product group. It provides a starting point for new mechanics without history. Modeling the post-campaign dip: The size and duration of the dip are learned according to the campaign type and the stockability of the product. In the model, separate lagged variables are defined for the periods after the campaign so that the dip is explicitly reflected in the forecast. Method selection depends on the data situation. Regression or machine learning stands out when there is a large campaign history, control group comparison when control units can be set aside for the application, and the analog approach for new mechanics without history. In practice, methods are often used together and results are compared to validate each other. How Is a Campaign Forecast Built? Cleanse the history: Campaign weeks and the dips that follow should be separated from past sales so that a clean baseline demand series is obtained.Build a campaign library: For each campaign, the type, mechanic (discount rate, multi-buy, extra display), channel, duration, period, and realized net uplift should be recorded.Derive type-specific effect profiles: The intensity of the uplift, its duration, the post-campaign dip, and persistence should be determined separately for each campaign type.Keep the layers separate: It is recommended to store the baseline demand and campaign layers as separate versions in the system. Which layer is wrong can only be measured this way.Monitor channel stock: Campaign loadings to distributors and the traditional channel occur before the sale to the consumer. For this reason, the difference between sell-in and sell-out must be tracked separately. Why Is Post-Campaign Analysis Necessary? It is recommended to conduct an evaluation two to three weeks after the campaign ends. In this evaluation, the forecast gross uplift, pull-forward, and net incremental sales are compared with the actual values. The cause of the deviation (display not implemented, a competitor's concurrent campaign, stockout) should be recorded in the library. In a campaign that experienced a stockout, the sales achieved do not reflect demand; the uplift of such campaigns should be marked in the library as "under-measured". The problem most frequently encountered in campaign forecasts in practice is systematic optimism. It is generally assumed that the campaign will reach its target, but results are rarely examined retrospectively. When the deviation and bias of the campaign layer are measured separately, this optimism becomes visible (see How to Measure and Interpret Forecast Error: Bias and Percentage Error Metrics). What really needs to be measured is not the total sales quantity in the campaign week; it is the net incremental sales that would not have occurred without the campaign How Is the Campaign Calendar Linked to S&OP? A situation frequently seen in practice is that campaign information reaches the production side after capacity has been planned. To prevent this, it is recommended that the campaign calendar be prepared at least as far ahead as the S&OP horizon and approved in the demand review meeting. In periods with capacity or material constraints, the timing of the campaign is evaluated as an option in the supply review. Shifting a campaign by a few weeks is often less costly than overtime or subcontracting. The options that can be used in campaign decisions are covered in the article Which Decision in Which Meeting in the S&OP Cycle? Calendar, Decision Logic, and Scenarios. Technology and Digitalization: Tools Used in Modeling Campaign Effects The result of a promotion does not depend on a single variable. Discount rate, display, catalog, seasonality, and competitor activity all have an effect at the same time. Evaluating so many variables manually is difficult. In modeling campaign effects, technology is used at different layers, from data infrastructure to scenario tools. Data Infrastructure and Promotion Library Modeling depends on data being collected correctly. Enterprise Resource Planning (ERP) systems are the source of price, promotion calendar, and shipment history. Sell-out and channel stock data received through point-of-sale data and Electronic Data Interchange (EDI) make the effect of the campaign reaching the consumer visible. Tools dedicated to trade promotion management, in turn, feed the campaign library by keeping the campaign's mechanic, budget, and calendar as structured records. Having every promotion fully recorded by product, customer, channel, and period is a prerequisite for modeling. Statistical and Econometric Models Regression, price elasticity models, and time series models with external variables (for example, autoregressive models to which a campaign indicator is added as an external variable) allow the campaign effect to be estimated in an explainable and interpretable way. Since the coefficients directly give numerically how much an increase in the discount rate raises demand, they are easily understood by the planner and the marketing team. Machine Learning Methods Machine Learning (ML) is the general name for methods that produce forecasts by learning patterns from data. Tree-based models such as random forest and gradient boosting, and artificial neural networks, can learn together the nonlinear and interacting effects of many variables such as price, display, calendar, weather, and channel. Because they can handle many products and campaigns under a single model, products with a limited campaign history can benefit from patterns learned from other products. Explainability techniques, which show why the model gave a particular forecast, are necessary for the result to be accepted by the sales and marketing team. Experiment Design and Causal Analysis The most reliable way of showing that the campaign really created the effect is a controlled experiment. The campaign is applied in selected test stores and not in similar control stores; the difference between the two groups is read as the effect of the campaign (A/B test, difference-in-differences). Where a campaign runs simultaneously in all stores, synthetic control methods can be used, which construct the no-campaign case from a weighted combination of similar units in the past. Scenario and Optimization Tools Once the campaign effect has been forecast, scenario tools are used for "what-if" analysis: net incremental sales, margin, and capacity load are compared by trying different discount rates, durations, or channel combinations. Advanced Planning and Scheduling (APS) systems and optimization methods can propose a feasible plan by evaluating the campaign calendar together with capacity and material constraints. The decision is made in a meeting attended jointly by sales, marketing, and operations. Real-Time Monitoring and Demand Sensing Demand Sensing is an approach that quickly reflects current signals such as point-of-sale, order, and external data in the short-term forecast. The forecast can be corrected with point-of-sale data from the first days of the campaign, and replenishment can be directed accordingly. Because smart shelf and cooler sensors within the Internet of Things (IoT) report shelf gaps and consumption instantly, campaign sales lost due to stockouts become visible. Generative Artificial Intelligence (GenAI), in turn, can prepare a draft of the post-campaign evaluation report and summarize the actual and forecast effect comparatively. At the same time, the model should not be expected to solve all problems. A promotion model cannot produce a forecast that exceeds the quality of the records it is fed with. If promotion mechanic, display, and duration information are kept incomplete, the model learns the effect incorrectly. For a campaign type that has not been tried before, the model provides only a starting point; its output needs to be evaluated together with the field knowledge of the sales and marketing team. The success of the model is limited by the quality of the campaign library; recording every promotion completely comes before leaving the forecast to the model Implementation: Checklist Clean past sales with campaign weeks and subsequent weeks flagged.Build the campaign library with type, mechanic, channel, duration, and realized net uplift information.Determine the intensity of uplift, its duration, and the post-campaign dip separately for each campaign type.Store the baseline demand and campaign layers as separate forecast versions.Carry out the net incremental sales analysis 2-3 weeks after the campaign ends and record the results in the library.Approve the campaign calendar in the S&OP demand review.Record every promotion digitally with type, mechanic, discount rate, display, and duration information; models can learn correctly only if these records exist.Where possible, verify the effect experimentally by separating test and control stores in campaigns. Related Articles Demand Planning and S&OP in Fast-Moving Consumer Goods (FMCG): 5 Critical TopicsHow to Measure and Interpret Forecast Error: Bias and Percentage Error MetricsWhich Decision in Which Meeting in the S&OP Cycle? Calendar, Decision Logic, and ScenariosStore-Level Replenishment and Inventory Planning in Retail Frequently Asked Questions How is promotion effect measured? Sales in the campaign week are compared with the baseline demand expected without the campaign; the difference is gross uplift. Subtracting the dip in the post-campaign weeks from this uplift gives net incremental sales. Where possible, a test and control store comparison confirms the result. Does a post-campaign dip occur in every campaign? If the product can be stockpiled at home and the discount is deep, a dip is observed in most cases. For perishable or hard-to-stockpile products, the pull-forward effect remains more limited. In campaigns such as sampling, a permanent increase in demand may be seen. Can the same method be used for every campaign type? The basic logic is the same; however, since the effect profile changes by type, parameters and sometimes the method differ. Regression and elasticity models are preferred for price campaigns, lagged-effect models for advertising campaigns, and the analog approach for new mechanics. Who is responsible for the campaign forecast? The owner of the campaign assumptions (mechanic, display, channel) is the sales and marketing unit. Translating these assumptions into demand and measuring the result is the responsibility of demand planning. When is artificial intelligence useful in promotion forecasting? It provides benefit when there are enough well-labeled historical promotion records. If records are incomplete, promotion data discipline must be established first. For new campaign types, model output should not be used alone and should be considered together with the sales team's assessment. References Blattberg, R. C., & Neslin, S. A. (1990). Sales Promotion: Concepts, Methods, and Strategies. Prentice Hall.Cooper, L. G., Baron, P., Levy, W., Swisher, M., & Gogos, P. (1999). PromoCast™: A new forecasting method for promotion planning. Marketing Science, 18(3), 301-316.Van Heerde, H. J., Leeflang, P. S. H., & Wittink, D. R. (2004). Decomposing the sales promotion bump with store data. Marketing Science, 23(3), 317-334.Gilliland, M. (2010). The Business Forecasting Deal. Wiley. Related Optimum Planlama Trainings finance_mode Demand Planning and Sales Forecasting Learn forecasting processes, methods, and analyses through hands-on applications. Detailed Information handshake Sales and Operations Planning, S&OP/IBP Build the balance of demand and supply with competitive tactics and consensus. Detailed Information package_2 Inventory Management and Optimization Classify inventory and optimize inventory quantities and costs. Detailed Information Institutions We Work With in Fast-Moving Consumer Goods Some of the institutions to which we provide training and/or consulting services in the fast-moving consumer goods sector are listed below. For all sectors, you can visit our References page. Akgıda, 2022 Alfarm, 2011 Anadolu Efes, 2019, 2021 Banvit, 2010 Barilla Gıda, 2011–2022 Baydöner, 2015 Bereket Döner, 2013 BTA Catering, 2010–2022 Cola Turka, 2016 Crown Bevcan, 2023 Duru Bulgur, 2022 Eker Süt Ürünleri, 2024, 2025 Elita Gıda, 2024, 2025 Essity, 2022 Eti Gıda, 2010 Flormar, 2017 Giz Kozmetik, 2014 Haribo, 2017 Hero Baby, 2015 Imperial Tobacco, 2016 İpek Kağıt, 2010–2015 Kellogg's, 2023 Kılıç Deniz Ürünleri, 2010, 2015 Lezita, 2017 Lidersan, 2016 Lila Kağıt, 2018–2024 Makbul Kuruyemiş, 2021, 2022 Migros, 2021 Milhans Gıda, 2011 Nivea Türkiye, 2019 Penguen Gıda, 2010 Penti, 2014 Rebul Kozmetik, 2021 Red Bull Türkiye, 2023, 2025 Sanat Ambalaj, 2023 Sunar Grup, 2025 Superfresh, 2013 Şölen Çikolata, 2010 Tamek Sagra, 2025 Ulduz Şokolad, 2018 Unifo Gıda, 2010 Wefood, 2026 Yemtar, 2010 Comments “Throughout the training, we had the opportunity to learn both the theory and the practical examples in a clear and fluent way. The content was up to date, comprehensive and directly usable in business life. The interactive delivery and real examples in particular made the training even more useful. Thank you.” Rabia Yıldırım, Planning Manager (Production Planning, 2026) “It was a very enjoyable training. It was an environment where I felt that new information was added to my knowledge.” E**** C****, Regional Sales Manager, Sunar Group (S&OP, 2025) “I would like to thank you for the General Planning and Scheduling in Production training I attended. The training was very productive in terms of both content and transmission and provided new perspectives. In particular, examples and practical explanations on capacity planning, material requirements planning (MRP), master production schedule (MPS) and scheduling techniques (e.g. line balancing, priority rules) helped me reinforce the subject. I gained valuable information on how to identify the bottlenecks we encounter in our production processes and how to find solutions. In addition, the strengths of the training were the support of theoretical knowledge with practice, explanations made through sample scenarios and detailed answers to the questions of the participants. I would like to thank everyone who contributed and wish the continuation of such technical depth trainings.” G**** Ç****, Production Planning Assistant Specialist, Tamek Sagra (Production Planning, 2025) To view all testimonials, click here. Request a quote for training or consulting tailored to your organization. Request a Quote