Rust, Lemon and Zeithaml, in their 2004 Journal of Marketing Article "Return on Marketing: Using Customer Equity to Focus Marketing Strategy" have a neat solution. It's to make the dependent variable the customer. In other words, when making marketing investment decisions, companies should understand how each investment impact core "orthogonal perceptions" for each of its customer segments, and then trace the impact of these perceptions on customer behavior in the form of acquisition and retention. They suggest thinking of each customer relationship as a string of future actions--either they will or won't buy, at different amounts, through time. They do this using a mathematical technique called a Markov Chain.
- How do we understand the impact of marketing investments on perceptions? The marketer still needs to do research, in the form of recognition tracking through time, to understand this.
- How do we understand which core attitudes are truly important? Market research needs to be done, in the form of qualitative followed by quantitative research, to understand the constructs that truly drive acquisition and retention behavior. In the article, the authors tested questions on inertia, quality, price, convenience, ad awareness, information, corporate citizenship, community events, ethical standards, etc totally 17 questions. They used PCA (Principle Components Analysis) to whittle this list down to 11 constructs.
- Customer segmentation. Not all customers act the same way, and the more discrimination we are able to build into our model, the better. Ideally, we'd like a unified segmentation to drive our research here.
- True short-term goals. While this approach can theoretically handle all marketing activities, it doesn't describe how to handle true short-term actions. However, I think it's pretty easy to bolt this on. Because we're looking at customers as the dependent variable, we'd simply add on known short-term acquisition and retention levers to only impact the first step of the Markov Chain. If an email acquisition program drove 5% of a segment to but, we'd simply add 5% onto the first Markov chain probability.
There's a lot more detail to this, but it should give a taste to the marketer who's interested in using customer equity as a dependent variable for ROMI / mix modeling. If there's more interest, give me a shout, go to MarketBridge and register your name, or download the Journal of Marketing Article (or all three).