Instead of weighting transactions (which I've done many times), one can build a regression model that predicts twelve-month top-line volume. An equation allows us to compare the relative value of File Power attributes.
I typically build four Logistic Regression response models, and four Ordinary Least Squares Regression spend models.
- Logistic Regression Response Models = 1x Buyers, 2x Buyers, 3x-5x Buyers, 6x+ Buyers, within 12 month file, 36 months of purchase history.
- Ordinary Least Squares Regression Spend Models = 1x Buyers, 2x Buyers, 3x-5x Buyers, 6x+ Buyers, within 12 month file, 36 months of purchase history.
I use classic RFM variables (square root of months since last purchase, frequency, average order value, historical demand used in spend models) ... and I use 1/0 indicators for key channel-based attributes and key merchandise categories.
The result? A measure of File Power for each of your twelve-month buyers.
Tomorrow? I'll show you how one can use File Power to quantify the impact of a marketing and/or merchandising strategy.