Seasonality is the most predictable part of most business series, because it repeats on a known calendar.
Spotting it stops ordinary seasonal swings from being mistaken for real changes.
Daily, weekly and yearly patterns are the most common periods.
Seasonal patterns can be additive, adding the same amount each season, or multiplicative, scaling with the level.
Multiplicative seasonality grows with the series, as when December sales are always 50% above average.
Taking logarithms turns multiplicative seasonality into additive, which many methods prefer.
Holidays that move, such as Easter, shift seasonal peaks between months and need special handling.
Calendar effects, such as months with 5 weekends, also change totals without any real change in demand.
Seasonally adjusted figures remove the expected seasonal effect so that underlying changes stand out.
Cycles differ from seasons because their length varies, which makes them much harder to forecast.
Business cycles lasting several years are a common example, as are the life cycles of products.
Some apparent cycles are just noise, since random series can look wavy over short stretches.
Detecting seasonality needs at least 2 full periods, so yearly patterns need 2 or more years of data.
In practice, comparing each period with the same period a year earlier is the simplest seasonal check.
The usual mistake is calling any repeating wave seasonal, when only a fixed period makes a pattern seasonal.