Seasonality and monthly forecast calculator
Most monthly forecasts are last year plus a growth guess. This tool does the same thing properly: it fits a trend to your history, works out how far each calendar month sits above or below it, and projects the next twelve months as trend times index. Every step is visible, so the forecast can be argued with rather than believed.
+4.8% on the last twelve months.
- History
- Forecast
- Trend × index
- Trend
- +5.1% a year
- Fit error (MAPE)
- 0.0%
- Last 12 months
- 772,500
- Months observed
- 12 of 12
- Highest month
- Oct
- Lowest month
- Feb
+256 contacts a month.
How far the fit sits from the history, on average.
Index 1.16.
Index 0.90.
Show the working
- 24 months of history from Jan. A straight line fitted by least squares gives a trend of 59868 at the start rising 256 a month.
- Each month's volume ÷ its trend value is a ratio; averaging the ratios for each calendar month and normalising to a mean of 1 gives the index: Jan 0.97, Feb 0.90, Mar 1.02, Apr 1.05, May 1.00, Jun 0.94, Jul 0.90, Aug 0.91, Sep 1.07, Oct 1.16, Nov 1.12, Dec 0.97.
- The fit is trend × index; its mean absolute percentage error over the history is 0.0%.
- The next 12 months extend the trend and apply the index: Jan 64,124, Feb 59,621, Mar 68,112, Apr 69,954, May 66,963, Jun 62,955, Jul 60,928, Aug 61,697, Sep 72,904, Oct 78,915, Nov 76,795, Dec 66,454; 809,421 in total.
This is the annual shape. Pebble WFM forecasts every interval of every day, with weekday and holiday effects. Free month, no card needed.
| Month | Index | Forecast |
|---|---|---|
| Jan | 0.97 | 64,124 |
| Feb | 0.90 | 59,621 |
| Mar | 1.02 | 68,112 |
| Apr | 1.05 | 69,954 |
| May | 1.00 | 66,963 |
| Jun | 0.94 | 62,955 |
| Jul | 0.90 | 60,928 |
| Aug | 0.91 | 61,697 |
| Sep | 1.07 | 72,904 |
| Oct | 1.16 | 78,915 |
| Nov | 1.12 | 76,795 |
| Dec | 0.97 | 66,454 |
How the seasonal index is calculated
A straight line is fitted through the history by least squares: that is the trend, and its slope is the underlying growth or decline per month. Each month's volume divided by the trend value for that month is its seasonal ratio; averaging the ratios for each calendar month across the years you have gives the index, normalised so the twelve average to one.
The fit is the trend times the index, and its error against the history is reported as a mean absolute percentage error, which is also a fair guess at how wrong next year will be if nothing changes. The forecast is the same trend, extended, times the same indices.
Two years of history gives each month two observations; one year gives one, which means the index cannot separate seasonality from noise. Three years is better. Anything with a step change in it, a product launch or a lost contract, should be trimmed to the period after the change.
Frequently asked questions
- How much history do I need?
- Twelve months is the minimum for an index; twenty-four is where it becomes trustworthy; thirty-six is ideal. With less than a year the tool will still fit a trend, but the indices for months it has not seen are set to one.
- What about public holidays and events?
- They are inside the monthly totals, so the index carries them if they fall in the same month each year. Easter moving between March and April is the classic problem; adjust those two months by hand, or plan them at weekly resolution.
- Is this better than the forecasting in a WFM tool?
- No. It is the simplest method that is defensible, and it is transparent. A workforce management tool forecasts at interval level with day-of-week and holiday effects, tracks its own accuracy, and reforecasts as actuals arrive. This tool is for the annual budget conversation.
Stop doing this one interval at a time
Pebble WFM forecasts your demand, computes the staffing requirement for every interval, builds the roster and publishes it, with self-service for agents and a copilot that can do what a planner can. Explore a sample organisation on day one.