### Calculating Jennifer

This Jennifer character, she's an interesting one.  Let's look at a few example of what Jennifer looks like.

Remember, I assigned the following weights by channel (your mileage may vary):
• Mail Orders = Weight of 0.00.
• Telephone Orders = Weight of 0.15.
• Online Orders Matched Back to a Catalog = Weight of 0.30.
• Search Orders Matched Back to a Catalog = Weight of 0.40.
• Email Orders Matched Back to a Catalog = Weight of 0.50.
• Pure Search Orders = Weight of 0.60.
• Pure Email Orders = Weight of 0.70.
• Online Advertising Orders, No Offline Interaction = Weight of 0.75.
• Pure Online Orders = Weight of 0.80.
• Mobile, Social, Flash Sales Orders = Weight of 1.00.
And remember, we weight historical transactions by recency ...1.00 for 0-12 month orders, 0.50 for 13-24 month orders, 0.25 for 25-36 month orders, 0.15 for 37-48 month orders, 0.10 for 49-60 month orders, and 0.05 for 61+ month orders.

So here's an example.  A customer spends \$100 60 months ago via telephone, and then spends \$100 today via a pure email order.
• Net Weighted Outcome = (\$100 * 0.10 * 0.15 + \$100 * 1.00 * 0.70) / (\$100 * 0.10 + \$100 *  1.00) = 0.65.
Remember, we categorize a customer based on the net weighted outcome:
• Judy = 0.000 to 0.333.
• Jennifer = 0.334 to 0.667.
• Jasmine = 0.668 to 1.000.
In this case, the customer is "Jennifer".  Her telephone purchase five years ago holds her back from being a "Jasmine", just barely.  As that transaction ages, Jennifer will become Jasmine.  Or, if she buys via email again, she'll become a Jasmine.

Here's another example:  The customer purchased online after receiving a catalog 13-24 months ago, and the customer purchased off of her iPhone yesterday.  The catalog-based transaction was \$200, the iPhone purchase was \$50.
• Net Weighted Outcome = (200 * 0.50 * 0.30 + 50 * 1.00 * 1.00) / (200 * 0.50 + 50 * 1.00) = 0.533.
This customer is "Jennifer" because her catalog-based transaction was four times as big as the iPhone transaction.  Again, as that catalog-based transaction ages, Jennifer is more likely to become Jasmine, especially if she purchases this way in the future.

Better yet, link your web analytics data to purchase data across channels ... account for Jennifer-like referring URLs in your weighting scheme.

I know, I know, you're going to have issues with this.  Well, you are welcome to do your own research, and come up with your own weighting scheme, I'd welcome it.  Heck, if you do it, I'll consider publishing it for all to see!

Tomorrow, we "calculate Jasmine".