gavinshearer.com

« marsEdit Rocks! | Main | Storm Beat Silver Stars »

May 29, 2005

Data Analyzin'

After an exceptionally pleasant run (and coffee at Zoka) with Jeff this morning, I returned home, grabbed a shower, and parked my butt in front of the Vaio for some analysis work on my Database Marketing project. I'm going to be here most of the day.

I know that only stats geeks are supposed to like SPSS, but, as a marketing person, I find formulating the "right" query akin to learning the secret words that open Ali Baba's cave - magical, and not just a little bit tingly-excited feeling. Jo and I have invested a lot of time scrubbing, re-coding and summarizing our data sources, which is paying off by allowing us to quickly, easily and (hopefully) accurately answer some substantive questions about our clients' business.

Let me give you an example. One maxim of marketing is 'know thy customer' so I started my analysis by asking, "OK, who's buying?"

A quick flick of the wrist to an SPSS menu (Analyzye -> Descriptive Statistics -> Frequencies), a drag-and-drop of the relevant variables (rc_gender, rc_state, rc_country, has_title) and I can quickly see:

• What percentage of the customer base is female (82.4%);
• Where they live (more than 50% of our customers live in just 8 states);
• What country they're coming from (97.6% in the United States);
• Whether or not they have supplied a "title" in their billing information (we're using this as a proxy for a business purchase. 86.2% are shopping for personal reasons)

These statistics are tantalizing. Now I want to know how the customer base breaks down based on their number of orders (how many people just ordered once? twice? more than twice?), as well as the average revenue per customer across the customer base. Then I can take those two pieces of data and rank the customer base into deciles, with the highest-spending customers in the first decile, the second-highest group in the second, and so on.

Now what do we see?

It gets better. How well, for example, does our ability to predict a repeat order from a given customer correlate with their presence on our mailing list? Or their state?

This stuff is truly, truly fascinating. I wish our data source were better (richer), but we've got enough here to make some good marketing recommendations. Our presentation is Wednesday.

(What a great way to end graduate school!)

Posted by Gavin Shearer at May 29, 2005 2:55 PM. Posted to UW MBA.

« marsEdit Rocks! | Main | Storm Beat Silver Stars »