Sales Forecasting for SMEs in Digital Ecosystems: A Comparative Evaluation of ARIMA and Random Forest Models

Sales forecasting helps businesses manage inventory and plan operations. In Indonesia, many Small and Medium Enterprises (SMEs) still rely on manual records. Moving to digital tools is tough, mostly because of limited access to data specialists and technical systems. This study looked at two forecasting methods, ARIMA and Random Forest, using actual sales data from a small electric shop. The focus was on the store’s top-selling item to keep the model comparison clear and grounded. ARIMA, a classic time series model, works well for short-term trends when the data is relatively stable. Random Forest, on the other hand, handles complex, non-linear patterns without needing the data to follow strict rules. ARIMA gave more consistent and accurate daily forecasts. For small businesses with irregular data and few resources, this model made more sense. While machine learning tools like Random Forest are powerful, noisy or incomplete datasets can still present challenges. In this case, better performance was achieved by a simpler approach. Traditional models are often overlooked in favor of newer algorithms, but this comparison demonstrates that their value is still evident—especially for SMEs, where tools need to be understandable and usable without advanced training. Forecasting doesn't have to be complicated to be useful. By applying methods that are easier to run and explain, smaller retailers can start making data- informed decisions. These findings point to a path for gradual digital adoption, using tools that match the conditions many SMEs face today.