A Great Predictive Metric: "Power"

You've probably heard about all of the geeky metrics that folks compute ... and you're probably saying "PRODUCE SOMETHING LESS NERDY, NOW!"

If you want something less geeky, something that tells you where your business is headed in the next year, calculate a metric called "Power".

Simply put, "Power" is the sales expected from your twelve-month buyer file in the next year.  Here's how you calculate it.

• Step 1:  Segment your 12-month buyer file however you wish.  We do this as of a year ago, say 2010.08.23.  Let's pretend that you have five segments ... A / B / C / D / F.  Let's pretend that you have 100,000 customers per segment.
• Step 2:  For each segment, calculate the average amount a customer in that segment spent from 2010.08.24 to 2011.08.23.  Let's pretend that As spent $100, Bs spent $50, Cs spent $30, Ds spent $20, and Fs spent $10.
• Step 3:  Count the number of customers in each segment as of today.  Let's pretend that today you have 80,000 As, 100,000 Bs, 120,000 Cs, 120,000 Ds, and 120,000 Fs.
• Step 4:  Multiply last year's value by this year's file counts, yielding file "Power"!

At this time last year, you had 500,000 customers who generated 100,000*100 + 100,000*50 + 100,000*30 + 100,000*20 + 100,000*10 = $21,000,000.  At this time last year, your twelve-month buyer file was capable of $2,100,000 of "Power".

As of today, you have 540,000 customers who are expected to generate 80,000*100 + 100,000*50 + 120,000* 30 + 120,000*20 + 120,000*10 = $20,200,000.  You have more customers, however, your customers aren't capable of generating as much "Power" as customers were capable of generating last year.

This is a particularly important concept for online / e-commerce folks, because your web analytics tools make it really hard for you to see how powerful your customer file is.

Best of all, this metric is predictive in nature, it doesn't tell you what happened in the past, it tells you what is likely to happen in the future.

Retailers tend to run this metric on a monthly basis (some weekly), so that they can understand key inflection points.

Power --- a metric you need to calculate!!

Value Grids and Lifetime Value

You probably already have something like this posted to your office/cubicle, right?

The "Value Grid" is a table that illustrates how much twelve-month profit you will generate from a customer with various Recency/Frequency attributes.

Freeze your file as of August 22, 2010.  Segment customers into Recency/Frequency combinations.  Then measure customer profitability across these segments, from August 23, 2010 to August 22, 2011.

Your benchmark is the Recency = 1 / Frequency = 1 segment.  This is how much profit you generate in the next year by acquiring a new customer.  If you lose $22.00 profit acquiring a customer, then you've got problems, because in this table, the customer pays back $6.52 in the next year.  Oh boy!

Similarly, you explore the cost to reactivate a customer against future payback.  If you have a 36 month 3x buyer with $2.43 future value, you might be willing to spend a few extra dollars to convert the customer to a 4x buyer.

Then look at the customers who pay the bills!  In this case, customers who purchased recently and purchased five or more times generate a boatload of profit, don't they? 

Create a Value Grid.  Post it on the wall of your office/cubicle.

Simple Tip: Customer Value By Day Of Week

Those of you who enjoy measuring long-term value might want to research customer behavior according to the day of week the customer purchases from your brand.

Retail customers purchasing on Saturday or Sunday have different future value than customers who purchase on a weekday afternoon or evening.

Online customers purchasing early in the week have different future value than customers who purchase evenings or weekends.

Catalog customers who buy during the in-home week have different future value than customers who buy two months after a catalog was mailed.

Those of you who analyze online visitation behavior will observe unique trends, based on the day/time the user visits your website.

Give it a try!