If you only have a small amount of monthly demand data, I’d avoid jumping straight into complex forecasting models. In my experience, methods like Simple Exponential Smoothing or Croston’s Method (for intermittent demand) tend to perform better because they don’t require years of historical data to be useful.

I’d also compare your forecast against a simple moving average as a baseline. Surprisingly, the simpler model often performs just as well when data is limited. As you collect more history, you can start experimenting with ARIMA or machine learning approaches, but with sparse data, keeping it simple usually gives more reliable results.