There are many approaches to seasonality in Causal. The following two are the most used, but feel free to take an approach that best fits your needs. Both approaches use the concept of relative time so familiarize yourself with it if you have not already.

Seasonality as a % of annual total

  • In this approach you can create a Seasonality variable that applies a % of the annual value over 12 months (7 days, 30 days, etc) so that adding Jan-Dec would add up to 1 or 100%
  • Make sure to set up the variable as relative time

  • Then use another input variable for your Annual Target
  • The resulting formula below uses "month - 1" as the time modifier. The -1 aligns Causal's time step index (which starts at 0), with Causal's month index (1-12 for Jan-Dec). Note the -1 is used because this model starts in January, if the model was to start in February a -2 should applied and so on for other Month start dates.

Seasonality as a variance to the average

  • The previous approach (% of annual total) divides an annual target throughout the year, whereas this approach adjusts monthly numbers for seasonality.
  • The Seasonality input in this case, should equal 100% for “average” months and 120% or 90% for above or below average months, respectively. (versus adding up to 100% for the year in the previous approach).