Estimating e-resource usage through time series modeling: The case of the GNAT Library

Significant increases in the costs of e-resources in recent years cause libraries to face uncertainties regarding budget management, and thus many low-budget libraries to terminate their subscriptions. For a cost-effective collection and budget management, uncertainty regarding user demands must be eliminated and predictions must be based on objective basics by utilizing rational methods. Time series analysis is one of the widely used methods in demand forecasting. The aim of this study was to estimate e-resource usage in libraries through time series analysis. In this context, usage forecasting was conducted through ARIMA modeling on usage statistics of e-resources subscribed by the GNAT Library. However, the results obtained by the study indicated that applied model was not able to accurately estimate usage. In 2023 estimation for full-text download statistics on Political Science Complete database between 2019-2022, the result for January were more consistent. When the most successful results in 2023 estimation based on journal usage statistics of OECD iLibrary database between 2016-2022 were obtained in May and December, and for book usage statistics in December. In the quarterly estimations for journal usage statistics, the results for the first quarter of 2023 were found to be more consistent than the three other quarters, but none of the quarterly estimates for book usage were consistent. In 2021 estimation with book section usage data of OECD iLibrary database between 2016-2020, the most accurate estimate was achieved in October. In the quarterly analysis of book section usage, the most accurate result was belonged to the third quarter of 2021. It is recommended that multivariate analyses be considered in future research. The application of univariate analyses without measuring the effects of different variables on e-resource use is seen as an important limitation. It is thought that usage estimates, by the same observation data, with other methods such as trend analysis might be more successful in estimating e-resource usage.