The premise of forecasting seasonality is to predict future trends based on historical datasets. Whether you’re predicting demand for the holiday season, fulfillment orders, same-store sales, or customer walk-ins for your restaurant, forecasting industry seasonality is vital to optimal profitability.
Case Study: Online Advertising
Within the ad tech industry there are constant factors which can affect profitability such as variability in bid requests, technical bugs, campaign performance, or algorithmic functions. Given the plethora of variables, forecasting day-to-day metrics is crucial to operating at an optimal level.
The graph shown is an example of forecasting future advertising spend per day based on historical data. The first step is to gather and analyze historical data in order to solve for day-to-day growth rates; the second step is to forecast the month’s total ad spend. The two analyses will provide an accurate forecast for each day’s spend for the current month. Given a highly seasonal industry such as online advertising, the dataset is the current month of last year’s advertising spend. To reiterate, the day-to-day growth rates are the only metric analyzed and applied to the current month’s forecast. The total month of current advertising spend is and should be a separate forecast based on current business trends.
Now that daily spend is forecasted for the month, all to do is monitor actuals are in-line with the forecast. Depicted in the chart, you can see how actuals are well within the forecast until mid-month. In this particular case, there was an internal bug, which was causing the algorithm to not properly win bids with particular ad exchanges. Because a proper forecast was set and monitored, the variance was red flagged, brought to the developers and fixed with a couple of days, and actuals were then realigned to the daily forecast. If a proper seasonal forecast was not set, irregularity in spend may not have been caught so quickly, which would have caused spend to drop for a longer period of time and subsequently reduce future profits.