Using Finite State Machines to Manage Customer Relations

If today’s marketing is supposed to be all about customer management, why are the predominant products called “campaign management” solutions? While consultants and analysts emphasize one-to-one optimized marketing, the available tools focus primarily on executing discrete campaigns.

What’s missing from traditional approaches is the ability to manage a longitudinal stream of interactions with customers recognizing that customer management is a process, not a disconnected set of communications. It is no surprise that response rates have fallen in many areas as technology has made it easier to blast out more campaigns with little regard for long-term effects.

Fortunately, database marketing technology is finally catching up to the real-world need to manage customer relationships over time. For example, finite state machines are being adapted to marketing and provide an entire new set of capabilities.

Finite state machine models have been used by computer scientists to model and manage complex processes in areas such as logistics and manufacturing for many decades. More recently, a specification for state machines has been incorporated into the Unified Modeling Language (UML) specification, the de facto standard for modern software development.

Despite their techie name, state machines are at heart fairly easy to understand.

Let’s take a simplified version of an onboarding program we’ve designed for one of our clients. In this program, we send new customers a welcome message that thanks him or her for their business and invites them back. Those that purchase again within 45 days then receive a follow-up “educational” mailing that provides more information about their current services and suggests complementary products and services, as well as customer service contact information.

Customers who do not respond to the welcome package receive a different educational package; if they still do not purchase, they will receive up to two promotional offers.

To create a state machine, we take each one of these stages, and model it as a state. For example, all new customers will be qualified into the initial state for this program. From the initial state, customers then automatically transition to the “welcome state,” where they remain for 45 days or until they purchase.

For each state, we then add transition rules, e.g., if a customer responds to our welcome message, move her to the education state. These rules will be implemented in software, so subsequent marketing actions will be highly automated.

Most of today’s campaign management applications could probably handle this level of complexity. But because of the modular approach and the ability to build hierarchical programs, state machines permit us to manage much more complicated processes.

On the downside, state machines cannot be used for true optimization. They can be used to model and manage complex marketing programs, but being rules-based systems, they require more optimum strategies to be discovered by outside means.

Even with this limitation, state machines remain a significant step beyond most of today’s campaign-centered solutions. They allow planning and execution to span relatively long periods of the customer lifecycle. For example, one manufacturer’s customer management program incorporates a long-term contact strategy, from an initial welcome communication for newly registered customers to successive follow-ups during the next several years.

David King is CEO of New York-based marketing solutions provider Fulcrum.