Kevin Hillstrom: MineThatData

Exploring How Customers Interact With Advertising, Products, Brands, and Channels, using Multichannel Forensics.

July 31, 2008

Starbucks Database Marketing

It wasn't that long ago that marketers lauded Starbucks, praising decisions like selling CDs and the like.

And then things happened. The business posts a quarterly loss for the first time in sixteen years, and the critics get to have their day while the marketing fans shy away.

Folks who work in Database Marketing departments like to diagnose the reasons why a business is struggling, from a customer standpoint. And when a brand is in Retention Mode, like Starbucks is, folks might look at a very simple set of metrics.

A typical dashboard might look at customers who shopped in a Starbucks store in June, measuring how likely the customers were to repurchase in July. The dashboard compares 2008 performance to similar customers in 2007. Of course, ninety percent of folks tell you "you don't capture every single customer, what about all of those customers paying with cash, customers you cannot track?" Those pundits would rather you do nothing than at least gaining some insight. I vote for gaining some insight.

This is what the a dashboard might look like within a specific market (the data are not representative of Starbucks, and are only presented for illustrative purposes).


Jul-08 Jul-07 ----Change




June Households 10,000 10,500 -4.8%
Repurchase Rate 0.782 0.831 -5.9%
Total Buyers 7,820 8,726 -10.4%
Orders per Buyer 8.341 8.492 -1.8%
Items per Order 1.243 1.184 5.0%
Price per Item 3.894 3.790 2.7%
Average Order Value 4.840 4.487 7.9%
Total Volume $315,713 $332,500 -5.0%




All Other Buyers 3,941 4,782 -17.6%
Orders per Buyer 3.743 3.593 4.2%
Items per Order 1.143 1.074 6.4%
Price per Item 3.631 3.504 3.6%
Average Order Value 4.150 3.763 10.3%
Total Volume $61,221 $64,660 -5.3%

Assuming you're able to capture the performance of some customers, you get the opportunity to get some level of insight. In this hypothetical case, June customers are less likely to visit the store, but are buying more items and spending more. This kind of simulates the "less traffic" situation you hear about.

Other customers exhibit a similar trend.

More often than not, pundits rip database marketers for making conclusions about customer behavior when not every transaction can be measured. Always remember to clearly state the limitations in the data being presented, but always remember that it is better to present something than to hide behind the crutch that you're not tracking cash customers. And web analytics pros --- you fall into the same boat!!

In no way am I saying that Starbucks does this ... I'm just using their brand for illustrative purposes.

Another tip --- when closing stores, you make strategic choices based on the customer base supporting stores. For instance, given two stores that are equally unprofitable, you kill the store that competes with more stores in the same geographic area. That's kind of a "duh", but, the data exists to make these kind of insights, so you may as well use the data to your advantage. Back in the day at Eddie Bauer, we evaluated every store based on estimated cannibalization rates --- understanding how each store was impacted by neighboring stores.

Labels: ,

2 Comments:

At 9:24 PM , Anonymous Bazily said...

Funny, was just discussing hwo Starbucks doesn't seem to use this hypothetical data for any kind of database marketing.

Just curious, what data are you talking about? We were thinking of the 12% of sales on gift cards...

 
At 10:33 PM , Blogger Kevin said...

It can be as simple as the sales on a credit card. All big retailers maintain a database of customers who use debit or credit cards --- not necessarily including the name/address, but recording each time that specific credit card is used.

That information alone is worth the effort to mine.

 

Post a Comment

Links to this post:

Create a Link

<< Home