Purpose: The study aims to analyze forecast errors for various time series generated by a 3PL logistics operator across ten distribution channels managed by the operator.

Design/methodology/approach: The research focused on ten distribution channels served by a 3PL logistics operator utilizing the Google Cloud AI forecasting tool as part of the Google Cloud AI service. The R environment was used in the study. The research centered on analyzing forecast error series, particularly decomposition analysis of the series, to identify trends and seasonality in forecast errors.

Findings: The analysis of forecast errors reveals diverse patterns and characteristics of errors across individual channels. Statistical tests for various channels show significant differences in forecast error groups in some cases, suggesting that the forecasting tool may perform more accurately for certain channels than others. A systematic component was observed in all analyzed Household Appliance Channels (seasonality in all channels, and no significant trend identified only in Channel 10). In contrast, significant trends were identified in one Pharmaceutical Channel (Channel 02), while no systematic components were detected in the remaining channels within this group.

Research limitations: Logistics operations typically depend on numerous variables, which may affect forecast accuracy. Additionally, the lack of information on the forecasting models, mechanisms (black box), and input data limits a comprehensive understanding of the sources of errors.

Value of the paper: The study highlights the valuable insights that can be derived from analyzing forecast errors in time series within the context of logistics operations. The findings underscore the need for a tailored forecasting approach for each channel, the importance of enhancing the forecasting tool, and the potential for improving forecast accuracy by focusing on trends and seasonality. This analysis makes a significant contribution to the theory and practice of demand forecasting by logistics operators in distribution networks. The research offers valuable contributions to ongoing efforts in demand forecasting by logistics operators.

Keywords: time series of forecasting errors, 3PL, logistics operator, demand forecasting, distribution channels,