The Financial Side of CRM
Budgeting-forecasting models are key diagnostic tools for management to decipher how changes in policies and decisions impact financial performance. Finance typically maintains these models, as their format is attuned to existing accounting systems and data. This natural fit makes it a well-established process that is used in all organizations that plan and evaluate performance via individual product/service categories.
In a CRM environment this process disintegrates. The assumptions and relationships on which conventional financial data sources and analytical methods are built become inadequate, because CRM requires a systems perspective instead of a component perspective. Instead of simply summing individual budgets and forecasts to create an enterprise version, the significance of CRM models lie in their ability to expose the effects of interactions within a company and a company its customers.
Key to creating a successful CRM model is finding the structure that provides the right balance between complexity, usability, and cost. For example, models with great numbers of fixed elements embedded in their logic are easy to employ; they require fewer user decisions. However, limiting choice decreases flexibility and hence the ability to explore divergent scenarios.
The same principle applies to scope. A model that forecasts financial performance strictly for one product, based on a single customer acquisition point, is fairly easy to create because you limit activities to those that are linear and consecutive--acquisition, retention, expiration. But the repercussion of this approach will be felt later. It prevents you from understanding the overall financial impact of your CRM strategy as the effects of interactions between product groups or business units in relation to customers are excluded.
Two basic types of CRM budgeting-forecasting models exist: activity-based and continuity-based.
Activity-based CRM models fit companies that continuously resell their goods; new orders require additional marketing activities. Accordingly, their projections are based on migrating customer segments through their lifecycle based on likely company or customer activities like promotions and orders.
Conversely, continuity-based companies sell an uninterrupted service (phone, utilities) or products delivered periodically (books, records, perishables) in return for regular customer payments. Their goal is not getting the next order, as customers have already signed up for the service, but retention. These models project results based on segmentation and consumption patterns in relation to time factors.
Fortunately, despite their differences, the prerequisites to develop these models are similar. They are:
A clearly defined objective:
Techniques to develop models depend on objectives, as objectives determine input, process, and output requirements. One company uses a model strictly to plan contact strategies for several interrelated marketing programs based on customer attrition and migration patterns; it has no interest in financial metrics. Another uses its model solely to understand the impact on financial results created by different customer acquisition and retention strategies.
An appropriate time horizon:
Weigh the implications of using time spans carefully. One- or two-year horizons are too short to evaluate the long-term effects of decisions. Conversely, long-term horizons require more assumptions and detail that can make it difficult to understand results and to operate the model.
Determine early the appropriate time intervals for input, process, and output. Most likely you will use some combination of weekly, monthly, quarterly, and yearly increments. As a rule, smaller increments provide greater flexibility but also require more research and analytical support.
Customer segmentation is the heart and soul of CRM, hence an indispensable ingredient of CRM budgeting-forecasting models. You have to be able to identify and understand how to create grouping of customers that are similar to each other, as the characteristics of those clusters are the frame around which all projections are constructed.
A customer migration method:
Financial results returned from conventional budgeting-forecasting models emulate anticipated product/services sales over time. CRM models, in contrast, recognize that those sales primarily reflect the quality of the customer base at a specific point in time. Hence, identifying the mechanics that migrate customers through their lifecycle while simultaneously capturing customer quality differences is quintessential. Preferably, the mechanics should expose both sides of cash flows--cost and revenue--and tie back to specific activities like promotions and orders.
Generally, you need three types of data: load, forecast, and validation. The first portrays your customer base at the model start-time, the second delivers information needed for the actual projection, and the third for its validation. All three are cultivated from historical information, so make sure it is available.
Having historical data is not mandatory, however. In fact, it can be an encumbrance if your company or its business environment changed dramatically in recent years. In such a situation use approximations. That way you can still explore and experiment with different scenarios and test the reasonableness or assumptions and expectations.
All the same, if you rely on historical data to provide the input to your model be aware of two major stumbling blocks: 1) combining data from multiple systems requires extensive data extraction and transformation; 2) established analytical methods and tools are woefully inadequate to support CRM models.
The reasons for both obstacles are that conventional systems, data structures, and analytical methods focus on the performance of individual business division or products. They are designed to support analysis and provide summarizations that roll-up information strictly within the constraints of the traditional chain of command tunnels. This is very different from the type of analysis needed in a CRM environment. The latter suggests that even if each function and process within a company works as best as it can, the company as a whole does not.
In fact, CRM analysis starts on the opposite end of conventional analysis. First, it focuses on the broad picture of how a company wants to serve its customers, then on the interactions between its individual business groups and their customers, and last on the individual components that make up those interactions. That way the analysis does not run astray of its primary goal, which is to make the most out of company-customer interactions through the optimization of interactions between business groups, as well as between business groups and customers as a whole.
Model Validation and Tuning
Tuning and validating go hand-in-hand. By matching inputs, and comparing outputs to your company's history over a specific time period you can evaluate the model's accuracy. Realistic results suggest a well-designed model; otherwise consider it a tuning opportunity that requires investigating and reconciling difference between actual and historical results. Designing a model is an iterative process. You learn and create anew from one version to the next by using the knowledge gained from the previous version as the platform on which to build the next.
Overall, realize that consistency is more important than accuracy, because erratic application of data or logic undermines the best intentions and certainly the faith of users. Start simple, with a broad idea, check if you are going the right direction, go back and redo with more detail.
Starting a CRM initiative is a big step for any company. It is expensive, time consuming, and risky, if published failure rates are credible. Constructing a CRM budgeting and forecasting model before the implementation can reduce that risk by illuminating gaps between the concept and its actualization; if you can not translate what your CRM initiative is to accomplish into at least a rough sketch model that can illustrate the advantages of your vision either the concept is incomplete, or it is complete but not worth pursuing.
Central to creating a CRM budgeting-forecasting model is the identification of relevant customer segments, interrelationships between business units, and the mechanics that actually move customers through their lifecycle. With that information in place the financial impact of customer activities become exposed, enabling a company to manage its customer base like an investment portfolio through the discovery of its own unique equilibrium between the need to acquire new customer and keeping existing ones.
1. Are the reasons for creating a CRM budgeting-forecasting model clearly defined?
2. Will expected users feel comfortable using the model?
3. Do the data elements needed to support the model exist?
4. Has the scope of the model and its limitation been clearly communicated to user?
5. Is the model balanced in regard to hard-coded logic and input requirements?
6. Does the model reflect as accurately as possible the interactions between business units and the company and its customers.
About the Author
Tom Richebacher is an information specialist with the EDS Business Intelligence team. His area of expertise is the creation of statistical and financial models based on database services that are used for customer relationship management purposes. He also develops the infrastructure and reporting systems needed for financial, marketing, and operational analysis and information delivery. Contact him at firstname.lastname@example.org