Using big data for demand forecasting and dynamic pricing

The research involved developing and implementing a big data-driven demand forecasting and dynamic pricing system for retail businesses in Ukraine. The methodology covered gathering a massive database from online trading platforms, transactional information from points of sale, and loyalty programmes, which showed an overall aggregated data quality index of 92.6%. Applying a comprehensive set of cleaning and normalisation methods boosted data quality by 12.47%. A comparative analysis of predictive models revealed the highest effectiveness of the LSTM network among individual models (R2 = 0.874, MAPE = 6.83%) and of the ensemble model among all tested approaches (R2 = 0.896, MAPE = 5.92%). Implementing the developed system in various retail formats, such as ATB-Market LLC, Foxtrot LLC, Nova Liniya PJSC, ALLO LLC, Silpo-Fud LLC, METRO Cash and Carry Ukraine LLC, Epicentr K LLC, Rozetka LLC, Comfy Trade LLC, and INTERTOP Ukraine LLC, showed a significant improvement in economic efficiency, with an average revenue increase of 9.16%, a marginal profit increase of 11.08%, and a 6.95% reduction in inventory levels. The best performance was demonstrated by the online stores Rozetka and ALLO, with Return on Investment figures of 516% and a payback period of 2.7 months. Regional analysis revealed significant differences in system implementation effectiveness, with the best results in the Western region, specifically Lviv and Volyn regions (revenue increase 9.27 ± 2.05%), and among non-food retailers, particularly Comfy Trade LLC and INTERTOP Ukraine LLC (9.86 ± 2.23%). Small businesses showed the highest adaptability to dynamic pricing, with a revenue increase of 10.23 ± 2.14% and an Return on Investment of 405 ± 86%. The research confirmed the high scalability and adaptability of the proposed approach for the Ukrainian market and allowed for the development of differentiated recommendations for system implementation for various types of retail businesses, considering their size, regional location, and product specialisation