If a client asks you to build a forecasting model with highly volatile historical data, how would you approach the problem?
