Data Driven: Responsive Forecasting
How do you create a response order curve to accurately reflect how close you will be to plan mid-way through a marketing campaign? How can that curve help efficiently manage staffing and/or inventory concerns? How do major holidays affect those forecasted curves?
Forecasting, by its very nature, leads to criticism. Unless you are 100 percent accurate every time out, you have "failed." However, reasonably accurate order response curve forecasts can allow an organization to efficiently manage staffing requirements during peak periods, solve potentially catastrophic inventory issues mid-season, and reasonably plan cash flow operations.
An order response curve is the historical calculation of orders—and percentage of orders—received weekly from a direct response campaign. Construct, from the past two years, a weekly response pattern of each promotional campaign. This is the percentage of orders received by week for the drop. If you have four mailings a year, you need to develop four weekly response curves, one for each season. If you have remailings of each major campaign, those need to be tracked and a separate response pattern built for each drop. Ideally, a direct marketer will have two or three years of history to compare for each seasonal drop.
From this information, you can identify the "half life" of each mailpiece or catalog—the point when 50 percent of the orders are in. The control buying team will use this "doubling point" when placing merchandise reorders.
A sample of an actual order response curve is shown in the chart. (Note: While the response order curves presented below deal with print campaigns, the same concepts hold true for either pure-play Web marketing campaigns or multichannel initiatives.)
Be forewarned, however. Even though a standard order curve is presented, you cannot assume that this response curve is appropriate for your company. You must build a response curve for your marketing campaign, one that incorporates your industry's nuances and seasonality. In addition, while order response curves are built upon historical data, major changes in product line, competitive landscape or economic outlook need to be factored in to maintain a reasonable level of accuracy.