Embedding vs. End Game
Good facilitators will find every opportunity to link a strategy discussion back to the customer. Who are they? What are they doing (or not doing)? How are we, as a brand, engaging (or not engaging)? What can we do better?
Even without assigning the term “modeling,” most marketers are (or should be) using attributes to discuss and estimate customer behavior and value. And this value informs decisions about strategy, tactics, measurement, and programs. Simply put, models and other statistically derived “information” are meant to derive additional knowledge about a product, consumer, product or any combination and therefore should be embedded in the marketing ecosystem. But even if you’re embedding modeling across channels and you’re leveraging statistical experience, there’s one key consideration that will make or break your technique: modeling has to be dynamic.
The one and done approach to modeling is short-sighted at best and ultimately doesn’t give you enough information to create and maintain a successful strategy. The result of modeling isn’t just a score. It’s a new attribute about customers and prospects –variables and values that are far from static.
So models have to be constantly maintained and verified and adapt as the customer evolves, which is happening faster every day due to channel adoption, merchandise, changes in marketing strategy, etc. You’re not getting full value out of your modeling exercise if you don’t make it dynamic and change it over time.
How many times have clients from fortune 500 firms RFP’d or called to say “we need a model for this project?” Serious statistical firms will walk away from this work knowing that six months or a year later, someone at the client will be using the inaccurate model for some unintended purpose. You have to be prepared to readdress this.
Modeling isn’t just for established businesses, either. Sure, you need some data to get started but even if you’re a new entrant to the marketplace, you can model buying behavior to know who is buying what products and how to impact response based on what you offer them. Legacy companies can dig a little deeper into their own data to explore things like transaction attributes and why their audience is becoming less responsive within specific channels. Just another example of why modeling is so important to the future of B2C.
Understanding these “modeled attributes” as they evolve gives marketers crucial insight to guide decisions around the customer experience, customer lifecycle, and the what and whens of promotion and optimization. And understanding your customer’s value to you helps you determine how best to present your value to them.
It’s important to remember that model-derived attributes should be pervasive throughout the whole marketing process. This means embedding that model score into the things you do – measurement, targeting, analysis, etc. And like anything else, this shouldn’t be done by silo – it’s throughout the entire business process you run, as well as the customer life cycle. So the next time someone in the office or cube says “we need a model,” ask why and how it can be used for the broader marketing effort.