Think of any real time series as being built up of trend, seasonal, cyclical and irregular parts. If you divide the actual sales figure for a period by the value that the trend line predicts for that same period, the trend part cancels out and what is left is the combined seasonal, cyclical and irregular effect. When this ratio is calculated for the same season across many years and averaged, the cyclical and irregular effects fade because they do not repeat with a fixed yearly pattern, but the seasonal effect keeps showing up regularly and survives the averaging. That is exactly why this ratio-to-trend approach is used to pull out the seasonal component, so the correct choice is Seasonal.
