Don’t Let Old Habits Dictate Your Marketing Thoughts
But sometimes it is hard to locate the right place to dig, due to limitations of experience (i.e., not knowing alternative ways are available or hidden), tool sets (i.e., wasting time with inappropriate tools for the job), or data (i.e., not having access to useful data, or having data in inadequate forms). If no one points out the correct path along the data journey, diligent workers would find alternate routes, often incorrectly or ineffectively, and that would become a set route for all to follow.
Recently we got a request from an apparel retailer to create summary reports of revenue dollars by major product categories for them. The query sounded like the requester was accustomed to using an item-level table in a traditional relational database. Now, if they are trying to rank or predict the most popular product lines for this or next season, sure, that is a good first step to take. What product line brought in the most cash?
But if the goal is to “sell” more of specific lines of products to their existing customers, looking at SKU-level product data one at a time isn’t the way to go, especially when such information is summarized in a form of the 360-degree customer view that marketers ever so desire these days (and in this case, such a customer-level summary has been newly created).
On an item level, what you have is a price of each item, which may or may not equate to a final payment amount on a transaction level (which requires some dealings with shipping, tax, discount, coupon and returns, too). Item price could be a proxy if you do NOT have a customer-level summary, but that is not the ideal variable to use when a customer-level summary is already available, showing how each customer spent their money on for each product line. In other words, work-arounds created in the past when such a view was not available should not dictate the way we look at the world indefinitely.
A product-centric view – set in motion by organizational habit – is quite inappropriate, if the goal is to find more buyers of a certain product line in the pool of existing customers. No one just buys from one product line forever, and it is the individual-level summary with multiple product buckets for each consumer that leads to insights; what combination of product and transaction behaviors lead to multi-category purchases? Adding up prices by product line won’t get you to a multi-dimensional view for effective cross-sell.
A while back, I had to deal with a similar situation with yet another apparel retailer chain in America. They had two major brands, though a buyer can purchase products of both brands in one physical store or website. Now, the assumption – which was validated first – was that multi-brand buyers would generate more loyalty and revenue in the long run. But to initiate a project like that, we needed proper customer-level tags that would show if each customer indeed bought multiple brands in her purchase history.
Unfortunately, every existing tag was wrong in the database, because it was either one tag on a customer level showing the last brand purchased (wrong!), or a transaction-level tag when multiple brands can be in one shopping basket (wrong again, and we couldn’t even figure out how anyone picked just one label for the transaction).
We had to build brand counters based on purchased items, in this case separate counters for Brand A and Brand B. Again, even in one transaction, both counters can go up. If you sum those up on an individual level, we can easily see who indeed touched multiple brands on a customer level. For example, if Brand A counter is positive, but Brand B counter remains zero, then the customer in question is a Brand A-only buyer. If both counters are positive, then she is one of the coveted multi-brand buyers.
Now, marketers can compare Brand A-only buyers, Brand B-only buyers, both brand buyers and all buyers, any which way they want, by a virtually unlimited number of variables, such as their purchase frequency, tenure, days from last purchase, average amount per transaction, average amount per customer, days between customers, not to mention any demographic variables like income, age, number of children, housing type, home ownership, etc. Just imagine how colorful such reports will be, and how easy it would be to obtain insights among these key groups. And if you want to find more multi-brand buyers from a pool of customers who only bought Brand A, you can easily create selection rules, or better yet, cross-sell models. As in, who are the potential buyers of Brand B among Brand A buyers?
Now, try to answer that question only with a dollar summary by product line. All you would get is what product generates the most cash (proxy of it, as you would be just counting “prices” in the item table, not knowing how many were returned later). Again, useful if you just want to know what product line has been kind to your business, but useless if you are trying to sell more of it to people who haven’t bought it yet, though they may have the “potential” to be multi-brand buyers.
The first step toward the right direction in this case, of course, is to establish proper problem statement. Is it about knowing the best product line of the past, or to acquire and maintain more multi-brand/multi-product line buyers?
This exercise gets to be fun if a company has more than two brands, and employs multiple promotion channels. Nevertheless, the way we would think about solving the problem wouldn’t be that different from the example that I provided here. If there are many brands, it would be a more complicated process, but never impossible to crack it down methodically. Just tell me which product line is the one that you want to push as a priority. (If you do not know which one to go after first, then that is a separate study altogether, necessitating creation of a different shape of datasets from the same transaction source.)
The bottom line is this: Please do not mindlessly go through the motions, the marketing thoughts, that have been carved into your organization and training modules. Always question the intent, and throw out any old methods and metrics if they do not serve your purpose today, no matter how popular they might have been so far.
If you don’t know what to do next, that is the right time to call in some outside help. And if you do, see if that consultant starts with the business-related questions first, or just wants to dig into the data immediately. Beware of the second type, as they may just stay in that mathematical comfort zone, charging by the hours for playing with data and wasting your money. If I may state the obvious one more time, we play with data to make money, not for preservation of old procedures or mathematical elegance.
Stephen H. Yu is a world-class database marketer. He has a proven track record in comprehensive strategic planning and tactical execution, effectively bridging the gap between the marketing and technology world with a balanced view obtained from more than 30 years of experience in best practices of database marketing. Currently, Yu is president and chief consultant at Willow Data Strategy. Previously, he was the head of analytics and insights at eClerx, and VP, Data Strategy & Analytics at Infogroup. Prior to that, Yu was the founding CTO of I-Behavior Inc., which pioneered the use of SKU-level behavioral data. “As a long-time data player with plenty of battle experiences, I would like to share my thoughts and knowledge that I obtained from being a bridge person between the marketing world and the technology world. In the end, data and analytics are just tools for decision-makers; let’s think about what we should be (or shouldn’t be) doing with them first. And the tools must be wielded properly to meet the goals, so let me share some useful tricks in database design, data refinement process and analytics.” Reach him at email@example.com.