Separating base demand and promotional effect in sales forecasting in the fast-food industry

This article examines a methodology for sales forecasting in the fast-food industry, based on separate modeling of base demand and promotional effects. The aim of the study is to improve forecast accuracy by isolating regular demand from the additional sales volume driven by promotional campaigns. The proposed approach involves preliminary filtering of promotional days from the time series, followed by the application of exponential smoothing (the Holt-Winters model) to estimate base sales. The difference between actual sales and the forecasted base series is interpreted as the promotional effect, allowing for a more objective assessment of the impact of promotions on final demand. The research methodology includes a comparative analysis of several strategies for processing promotional day data: no filtering, complete removal, and replacing values with typical indicators. Experiments conducted on data from a restaurant chain showed that the strategy of replacing sales on promotional days provides the best result, with a WAPE of approximately 16.7% compared to other approaches. The results indicate a reduction in the risk of product shortages during peak demand periods and an increase in the efficiency of inventory planning, which has important practical significance for optimizing operational activities in the fast-food sector.

1. Mkhitaryan S.V., Danchenok L.A. Sales Forecast Using Adaptive Statistical Methods. Fundamental Research. 2014;(9–4):818–822. (In Russ.).

2. Lukashevich N.S., Temirgaliyev E.R., Baranova T.V. Sales Projection Based on Model of Holt-Winter. Bulletin of Karaganda University. Economy Series. 2022;108(4):172–179. (In Russ.). https://doi.org/10.31489/2022ec4/172-179

3. Napolskaya Yu.V. Multifactor Regression Analysis and FMCG Retail Sales Forecasting in Offline Stores Using POS Data. Beneficium. 2024;(4):49–57. (In Russ.). https://doi.org/10.34680/BENEFICIUM.2024.4(53).49-57

4. Mikhailov D.V. Analiz praktiki primeneniya algoritmov mashinnogo obucheniya v rossiiskikh roznichnykh setyakh dlya razvitiya operatsionnoi deyatel'nosti i razrabotki marketingovoi strategii predpriyatiya. In: Fundamental and Applied Approaches to Solving Scientific Problems: sbornik nauchnykh statei po materialam XII Mezhdunarodnoi nauchno-prakticheskoi konferentsii, 28 April 2023, Ufa, Russia. Ufa: Nauchno-izdatel'skii tsentr "Vestnik nauki"; 2023. P. 122–141. (In Russ.).

5. Ustinov D., Emelyantsev D., Durov I., Tatarenkov A. Automation of Pricing in Retail Using Machine Learning. Universum: tekhnicheskie nauki. 2024;(6–1):20–26. (In Russ.). https://doi.org/10.32743/UniTech.2024.123.6.17762

6. Bazhanov N.N. Eksponentsial'noe sglazhivanie kak metod prognozirovaniya vremennykh ryadov. In: Teoriya i praktika nauki tret'ego tysyacheletiya: sbornik statei Mezhdunarodnoi nauchno-prakticheskoi konferentsii, 07 March 2014, Ufa, Russia. Ufa: Bashkir State University; 2014. P. 194–196. (In Russ.).

7. Pegachkova E.A., Yakovleva A.G. Analiz vliyaniya razlichnykh faktorov na prodazhi s pomoshch'yu algoritmov mashinnogo obucheniya. In: Materialy XXIII mezhdunarodnoi konferentsii po vychislitel'noi mekhanike i sovremennym prikladnym programmnym sistemam (VMSPPS'2023), 04–10 September 2023, Divnomorskoe, Krasnodar Krai, Russia. Moscow: Moscow Aviation Institute (National Research University); 2023. P. 127–128. (In Russ.).

8. Afanasieva T.V. Forecasting Time-Series Local Trends in the Big Data Analysis. Automation of Control Processes. 2014;(4):42–47. (In Russ.).

9. Bulgakov Yu.V. Adaptive Models Forecast of Sales. Journal of Marketing in Russia and Abroad. 2011;(4):3–14. (In Russ.).

10. Mkhitaryan S.V. Adaptivnaya trend-sezonnaya model' prognozirovaniya prodazh kompanii s uchetom predpolagaemykh izmenenii tempov rosta. Khroniki ob"edinennogo fonda elektronnykh resursov "Nauka i obrazovanie". 2016;(12):4. (In Russ.).