### Lifecycle Marketing and Touchpoints: Weighting

One of the biggest problems facing an analytics or marketing professional is the weighting of data.

Here's an example.  A customer purchases on August 1, 2008, purchasing on the website, spending \$100.  Then, on August 1, 2009, the customer purchases via paid search, spending \$100.  Finally, on August 1, 2010, the customer purchases via e-mail marketing, spending \$100.  Is this customer a loyal website buyer, a search buyer, an e-mail buyer, or a multi-channel buyer?  And how does the customer skew across channels?  Is the customer weighted equally, or do recent transactions carry more weight.

When I first obtain customer data, I like to run a quick regression model.  Any transaction in the past year represents the dependent variable.  My independent variables represent dollars spent in each prior year.  I'll obtain an equation that looks something like this, using just twelve-month buyers in the equation:
• 2011 Spend = 30 + 0.65*(2010 Spend) + 0.30*(2009 Spend) + 0.15*(2008 Spend) + 0.08*(2007 Spend) + 0.04*(2006 Spend).
Armed with this equation, I have a weighting scheme I can use to understand the impact of older transactions in the customer lifecycle.
• 12 Month Purchases = Weight of 1.000.
• 13-24 Month Purchases = Weight of (0.30 / 0.65) = 0.462.
• 25-36 Month Purchases = Weight of (0.15 / 0.65) = 0.231.
• 37-48 Month Purchases = Weight of (0.08 / 0.65) = 0.123.
• 49-60 Month Purchases = Weight of (0.04 / 0.65) = 0.062.
In my earlier example:
• August 1, 2008 Website Purchase Weight = \$100 * 0.231 = \$23.10.
• August 1, 2009 Search Purchase Weight = \$100 * 0.462 = \$46.20.
• August 1, 2010 E-Mail Purchase Weight = \$100 * 1.000 = \$100.00.
I sum weighted dollars:  \$100.00 + \$23.10 + \$46.20 = \$169.30.

Finally, I divide weighted dollars in each channel by total weighted dollars:
• Website = \$23.10 / \$169.30 = 13.6%.
• Search = \$46.20 / \$169.30 = 27.3%.
• E-Mail = \$100.00 / \$169.30 = 59.1%.
Given the timing of the purchases, and the weighting of each purchase, we can categorize this customer.  This customer is about 60% email, about a quarter search, and the remainder represents pure online purchases.

Does the weighting strategy make sense?  Weighting is important, because we do not want older transactions to skew the customer profile, and we do not want to focus only on new transactions.