Based on many years of marketing analytics experience, we have found several applications that are common and foundational across marketing organizations that understand their customers well and interact with them efficiently and effectively.
Customer Attributes Data Store
This is less an application and more a foundational data layer that makes analytics work; yet, it is missing from most marketing systems. This is the 360 degree customer-level view of activity and characteristics. For each customer, how frequently do they buy, how recent, how much, what types of products, what is their activity by channel, what are their demographics, etc. This will frequently be hundreds or even thousands of attributes for each customer that describes them in relation to the business.
For most analyses, the transactional data where many marketing systems leave off is analytically worthless; the detailed data has to be summarized and organized into a usable analytical form. To be clear, this isn’t a reporting table with customer population summaries by a few dimensions (e.g. time, geography, and product); it is far broader and is at the individual customer level. If you want to predict or analyze individual customer behavior, you need this data layer.
Customer Migration Reporting
This application provides the basic profile of the customer universe broken by meaningful enterprise segments (e.g. 12 month recency one time internet only buyers, 12 month recency 2+ time internet only buyers, etc.) with their activity over time periods (year to date, quarter, month, etc.) and comparisons to current and prior year. Combined with customer movement modeling, this provides the basic "health report" on your customer file – is your customer base growing or contracting, how are they performing relative to last year, how much are they buying, etc.
Subsequent Value Reporting
As mentioned earlier, one of the things you need to know is how much a customer is worth to you over time, and subsequently, how much you can afford to spend to acquire new customers. This is the application that answers that. Based on order triggers, it reports the subsequent activity, sales, profit, etc. for customers over a period of time after the trigger.
Campaign Reporting with Revenue Attribution
Most organizations have the ability to track incoming sales in an operational mode; this application tracks sales and profits back to the outgoing promotion and channel that generated them. This process is vital to understanding the success or failure of your marketing efforts. Revenue attribution is the process of attributing sales to the marketing efforts that generated them. Without appropriate revenue attribution, you mistakenly attribute too much or too little revenue to marketing channels and efforts; consequently, you don’t accurately allocate your marketing resources.
For example, if you send out a catalog but fail to account for the fact that the contact also drives internet revenues that can’t be directly tied to the catalog, you are underestimating the value of the catalog and will mistakenly believe that you can derive the same internet revenue with less catalog circulation. Similarly, if your paid search internet traffic is not properly attributed, you may believe that your organic web site revenues are greater than they actually are. The same issue applies to every channel and marketing effort.
Often, revenue attribution requires hold out tests to determine the cross channel impact. To provide reasonable readings, the tests are typically run for a relatively long time (6 months to a year). The end result is a set of business rules and algorithmic weightings that are either implemented directly as part of reporting applications or are applied to an Attributed Orders Data Store where the revenue data is stored in its final allocated form and is available for input to any application.
This is one of the more sophisticated marketing analytics applications. All the other applications we have talked about should be within the reach of most business users given the right infrastructure and tools; predictive modeling is more specialized. The Customer Attributes Data Store described earlier drives the modeling; the goal is to use that data in combination with known outcomes to predict customer behavior (who is likely to buy, remain a customer, leave, pay, etc.). While current modeling tools do not necessarily require trained statisticians, modeling does require special skills, but the potential benefits certainly warrant the required staffing or consulting resources.
Operationally, this is where all the analysis and knowledge is brought together to put the appropriate customers together with the right marketing promotions. If the organization has sophisticated predictive modeling in place, this may be a relatively simple process of using model scores in combination with other customer data to define the campaigns.
Beyond the foundational level, contact optimization replaces the typical marketing event centricity (choosing the best customers for an event) with customer centricity (choosing the best events for each customer). Contact optimization utilizes predictive sales per mailing customer scores, saturation measurements between marketing events, and optimization techniques to minimize contacts, maximize profitability, and enhance the customer experience with minimal revenue impact; it finds the optimal contact strategy for each customer across channels, brands, and time.
These applications are heavily interconnected; consequently, there are significant advantages to having them on a single platform. Once you have the appropriate platform and basic applications in place, a whole new world of opportunities, including sophisticated solutions like optimizing your customer contacts across the enterprise, presents itself. All of this moves you ever closer to the marketing "holy grail" of getting the right offer, to the right customer, at the right time, via the right channel.