Plan requirement
| Subscription | Any plan |
| Also required | Workforce Management |
| Access | Admin |
Which forecasting method to use, and why that choice matters considerably less than the quality of the data behind it.
What they differ on
How much weight recent weeks carry, how strongly seasonality is applied, and how quickly the forecast reacts to a change in trend.
All of them are reading the same history; they disagree about what to emphasise.
The rough guide
- Stable volume, clear weekly pattern. A method leaning on seasonality works well.
- Growing or shrinking steadily. One that weights recent data, or the trend is always behind.
- Volatile, campaign-driven. No algorithm handles this alone; you will be adjusting manually.
Test rather than reason about it
Run two against a period you already know and compare with what happened. Twenty minutes, and it answers the question definitively for your data.
Choosing on the description of the method is guessing.
Data quality beats algorithm choice
A clean history with holidays marked and abnormal weeks excluded, run through an ordinary method, beats messy history through a sophisticated one every time.
If the forecast is poor, look at the data before changing the method.
Do not change it often
Switching methods every time a week is off means you never learn whether any of them works. Choose one, run it for a quarter, and judge it on the quarter.
Per channel, possibly
Phone and email behave differently enough that different methods can suit them. Worth testing once rather than assuming one setting fits both.
Comments
0 comments
Article is closed for comments.