Please advise why or why not this methodology is correct. While I understand a lot of this is gut feeling, I would like a more quantifiable way to approach forecasting.
1. Forecast monthly call volumes based on past-years data using percent increase/decrease from one year to the next. Assumption is that calls received should hold a relatively similar pattern from one year to the next due to seasonal trends--holidays, seasons, etc.
2. Forecast weekly using using last years historical data for weekly calls received trends. Assumption is the Week1, Week2, Week3, Week4 will take the same trend cycle based on seasonality.
3. Forecast daily using historical data on daily trends--x #calls on Monday, Tuesday, Wednesday...Assumption: consumer behavior is unlikely to change over time--most calls will be received on Monday, for instance and the least amount of calls will arrive on, say Sunday.
4. Forecast the intra-day schedule based on historical time-of-day trends. Assumption: Same as the weekly. |