From source code to source path: understanding a customer's path to purchase
Since the beginning of direct mail time, marketers have been attempting to allocate specific purchases to specific efforts, placing key codes on catalogs, capturing them (as best they can) on orders and placing that source code (or a translated version) into the order record.
Five years ago, marketing was easier. The circulation plan was the path to achieve goals for the year and was developed and adhered to with fierce determination.
It is fair to say that consumers have moved quickly past the companies attempting to market to them in how they shop. Now, consumers can deploy every weapon at their disposal to buy the product they want, when they want, at the price they perceive as fair and with shipping terms they like, through a variety of purchasing channels.
Of course, we marketers are trying to catch up. A number of lifetime value studies have been done that can nail down comparative value of a shopper by channel. Does a search only buyer have more or less total LTV than a direct mail/search sourced buyer? Can we convert one-time buyers without direct mail instead of relying solely on e-mail? Is direct mail necessary to keep the really loyal customers loyal? And, if we develop facts to these and a myriad of other questions, at what rates are those facts changing, and directionally where are they going?
Without some reasonable data, we marketers have the potential to make decisions on budget allocations across channels with less than perfect data. This is particularly worrisome with mid-tier multichannel marketers, where pressures from rising costs and lowering prices are perhaps most fierce and the consequences of poor decisions can be felt immediately.
Over the past several years, companies have attempted - either internally or externally - to develop methodologies to allocate transactions to marketing efforts through a process known as matchback. How much of this transaction was due to the search, and how much to the direct mail piece? Generally, this matchback has had one best winner - one channel getting all of the credit for the purchase. And, in many instances, the results of this attribution aren't even recorded on the marketing database, but simply used for reporting purposes. When it is recorded, it usually overwrites the original effort. At its most sophisticated, the database will contain the original source code and the attributed source code.
To truly develop the necessary data to optimize marketing spend, I believe that a rules-based fractional promotion media-to-transaction allocation methodology contains the most promise. In this system, all potential marketing channels would be identified and their attributes measured. Relevant attributes include the order curve of the channel, the products contained in the message and other information about the offers. All of this information is metadata about the campaign and channel. Through analysis, rules can be developed to assign weights to those components to fractionally allocate the revenue of an order across those channels so that each promotion media receives the appropriate credit.
As an example, suppose that Company A had sent two catalogs to a customer, both with still open order curves, and had also sent an e-mail to the customer offering a discount on a certain product. Additionally, that e-mail was sent most recently, and its order curve was still open. If a customer clicks on the e-mail and purchases the specific product advertised in that e-mail, then the e-mail may possibly receive say 75 percent out of 100 percent of the order, with the most recent catalog receiving 20 percent and the older catalog receiving 5 percent. However, if that same customer came in through branded search and purchased a product from the first catalog, the weights may be shifted around entirely.
While it sounds complicated, rules-based fractional allocation can be made to work through rigorous analysis of your customer data and past promotions to develop a baseline for the rule set. And, through sensitivity analysis, rule sets can be tweaked to align more correctly with reality. And, the rule sets, metadata and allocations must be stored on the marketing database to continually monitor and refine results. Similarly, the fractional allocation engine must reside within the marketing database because such deep analysis would be impossible without a holistic view of both promotion and transaction data.
Solving the problem of correctly allocating revenue to marketing expense is taking on greater importance with the increasing disparity in expense between channels. However, blindly cutting direct mail circulation is not the right answer - getting the right marketing mix is.