Many forecasting methods are used to predict sales, such as the moving average method, naive method, exponential smoothing methods, Holt's linear method, and others. The results brought by these models are quite different. Forecast delivered by the naive method is entirely accurate for an extended period, like 3–5 years, Holt's methods are bringing accurate one-year period forecasts. The planning decisions have several levels, meaning different forecasting results. However, the authors that are testing various forecasting methods are not discussing results researched in different planning levels (retail chain and different pharmacies). The study is given to the construction of the forecasting model covering both planning levels, which later is empirically tested for the Lithuania retail case:
The development of the forecasting model for reduction of shortages in drug supply. To achieve this goal, the author revises the improvement of drug availability weekly.
The construction of the forecasting model is incorporating outliers' detection methods and sales by pharmacies to minimize shortage. In the forecasting model, the author uses Theil's U2 test to evaluate forecasting accuracy.
During analysis, the author constructs the model application for forecasting drug sales where weekly availability is highly recommended. The results show that forecasting on individual pharmacies level using the integration of these plans approach leads to higher accuracy.
The research covers 3 months of sales data. Das and Chaudhury suggest for short-sales period products use 36 days' time horizon. Ayati et al. discuss short and long-term time horizons for planning sales of drugs. Kanyalkar and Adil analyzed multi-site production and suggest that the time horizon should cover the longest lead time required for delivery of raw material, which is 12 weeks, and select 3 months (i.e., 13 weeks) as short-term time period horizon. Wongsunopparat and Chaveesuk forecast drug sales for 1-month and 12-month periods and compare the results. In this study, the focus is on short-term time-horizon, which is considered as 3 months period and also represents the longest lead-time. In the future, the study could review other periods. The author has incorporated the review of eight forecasting methods into the study by leaving other forecasting methods unresearched. Future studies could also incorporate different ARIMA methods into shortage reduction case analysis.
Presented forecasting model could be useful for practitioners, which analyze the reduction of the shortage of prescribed drugs. There the revision of repeated purchases is recommended for national authorities, wholesalers, and pharmacies aiming to minimize shortage.
The analysis to reach the highest forecast accuracy and identification of a forecasting approach which responds to the fluctuation of weekly sales for the whole pharmacy chain and separate pharmacies. The study contributes to drug sales review, where most authors analyze the total volume, which is not separated by pharmacies.