May 06, 2009

E-Mail Customers: They Can Be Valuable!

Part of any Multichannel Forensics project is an assessment of the value of e-mail subscribers.

Now this isn't some sort of qualitative "Brand 'X' uses great subject lines" or "E-Mail is a wonderful one-to-one relationship vehicle" analysis. Nope, we're actually controlling for key factors, understanding when all factors are equal if e-mail can carry the freight.

And it turns out that the fact a customer is an e-mail subscriber is important.

You'll see something like this:
  • Logistic Regression Of 12-Month Repurchase Rates = -2.000 - 0.5*(months since last purchase ^ 0.5) + 0.8*(number of orders) + 0.3*(email subscriber).
  • Spend Model = 120 + 0.30*(historical spend) + 4*(email subscriber).
So let's compare two "equal" customers, one an e-mail subscriber, one not. Each customer has a recency of 8 months, three orders, and $300 of historical spend. The first customer is an e-mail subscriber, the second customer is not an e-mail subscriber.
  • Customer #1: Response = 32.9%, Spend = $214, Value = $70.34.
  • Customer #2: Response = 26.6%, Spend = $210, Value = $55.89.
With two equal customers, e-mail is projected to deliver $14.45 of incremental sales in the next twelve months.

Your job is to compare this metric to the data you get via open rates and click through rates and conversion rates. Do you see a difference?

For instance, your forecasts might suggest that Customer #1 will generate $0.15 per e-mail, across 52 annual e-mail campaigns. If true, then you're forecasting this customer to generate $7.80 over the next twelve months.

In other words, the modeling procedure shows you that e-mail marketing is nearly twice as valuable as your traditional marketing metrics suggest. Complement this data with actual mail/holdout results, and you might really have something!

The important thing, of course, is to put your e-mail program through a full Multichannel Forensics analysis, thoroughly understanding the role e-mail plays in growing each channel.

No comments:

Post a Comment