4. Create 'Model-friendly' Behavior Predictors.
Once the partner who helped organize the data into an accessible structure is finished, it's time to extract a story from the data. The objective is to identify the key pieces of information about individual customers or the household that will help predict behavior.
How do we get there?
A skilled marketing data scientist with experience connecting large datasets—and sifting for gold—should be engaged. This experience is needed to work with database developers to uncover key pieces of information and plug them into analytic tools to provide a solution.
Together they will work with marketing to:
- Map each marketing outcome for all individual customers. This can be an extensive task; especially if you've collected several years' worth of transaction and marketing data.
- Build an analytics layer to help summarize data, such as how long you've had a customer, whether the customer purchased one product or many, how many contacts you've made, which contacts elicited a response, and whether your customers influence others positively or negatively.
5. Always Test and Learn.
These recommendations will not be a fast fix to an ailing marketing program. Instead, expect to take a phased approach for implementing these ideas. While doing so, it's best to always keep a control group to measure the incremental effect of each marketing program.
As programs become more complex, questions will arise that only a well-designed test can answer. Many statistical software solutions provide tools for the construction of well-designed, structured tests that will maximize learning with minimum cost.
Knowing customers and how they behave ranks first in importance for marketers. Collect this data with abandon, buy storage to save it and mine it like gold. Once internal behavior is understood, finding new customers who fit a "best behavior" profile, either online or offline, becomes a planned acquisition strategy.