Achieving a Confident Loyalty Program Launch
One That Leverages Pro Forma Modeling
One That Leverages Pro Forma Modeling
Loyalty programs need to continuosly evolve to keep up woth the changing customer and competitive landscape.
In order to drive increased acquisition, growth and retention, it is critical to offer a fresh and innovative program that will engage the member. A loyalty program typically has many levers to achieve this—the central earn/award structure, partnerships, and social media, to name a few.
When creating or perhaps completely relaunching a program, part of the process includes understanding the potential future return on investment and associated risks. An informed Economic Model can provide this insight and more. A model can range from simple to intricate, depending on program complexity and available data. It will evolve concurrently with other tracks such as concept development and customer research.
What do we know about the current customer base? A baseline needs to be established from which everything in the model builds. If we’re considering a first-time program, that might be a tall order. Transactional and other behavior at the identifiable customer level is not always available due to data capture and technology constraints. However, there is always some type of reporting or data available, often within the Finance group, which can provide directional input. Assumptions about total customers, average transactions and spend and retention rates are typical, required baseline metrics. If a program is currently in place, a much richer data source is likely available.
What do we believe the program will achieve in driving incremental behavior? By engaging a member through a loyalty program, we expect behavior change to come in different forms and to vary by channel and customer segment. The key is predicting what these changes may be how they may range. Assumptions are based on a combination of inputs. Insight from other companies where programs are closely measured should always be considered. Customer research provides insight that can be specifically aligned with the program concept. Measured results from CRM programs, marketing promotions or an existing program hit closest to home.
How much is this going to cost? Our primary goal is to estimate future return on investment. The potential program costs are the final input. There are both one-time build costs and ongoing costs. Technology, management, creative, communications, promotional, and analytics are the main categories. But let’s not forget reward cost. To what extent do we expect members to take advantage of rewards they have earned or been gifted—or to put it another way, what is the “breakage”?
Now we have the baseline, the behavior impact assumptions, and the cost. This is the primary flow of the model from top down, with the net impact and ROI as the final chapter. The next consideration is the time frame of the model. Are we modeling annually or monthly? How far into the future? Typically, the summarized model should forecast net impact five years into the future with an annual view. However, certain aspects of the model, enrollment for example, may drill down to a monthly level. Much of this depends on the business model. A high-frequency business might require a monthly forecast for certain components, where annual forecasts can be sufficient for a low-frequency or subscription-based business.
We know our behavioral impact assumptions are not exact—they might be too high or too low once reality plays out. Oracle Crystal Ball is a powerful Excel Add-In for predictive analytics, simulation, optimization and forecasting. This tool allows for transforming each of the assumptions into a distribution that can be fully defined by the modeler. For example, if our assumption is that retention will show a 2% improvement as a result of the program, it can be set at 1%, 2%, and 3% with equal likelihood. After setting all of the assumptions against a likelihood distribution, the model can be simulated thousands of times. The net impact and ROI, or any other outcome variable, will recalculate each time based on the individual assumption likelihoods. This provides a range of potential outcomes, from the most likely to least likely.
One last issue that should be addressed is the sensitivity of the assumptions. What assumption is most likely to raise havoc on the outcome if we get it wrong? The Crystal Ball simulation outputs a sensitivity analysis that will provide clues. For example, it might indicate that program enrollment drives the greatest variation in the potential net impact. If that is the case, we better get it right or have a plan in place to ensure acquisition is a key marketing focus.
There are many benefits to building a robust economic model. Being able to demonstrate a new or relaunched program’s potential contribution to the bottom line is critical to broad support and buy-in within the organization. Being able to inform the program concept and strategy, at launch and beyond, will only increase the likelihood of success. These types of models tend to be full of assumptions and any one of which could drastically alter the final ROI numbers. But if you can pick up clues about where your program should focus to drive the best probability of success, and then tailor your tactics to maximize results in that area, you will have both a solid business case and a more data-driven program in general.
Brad Davidson brings 20+ years of experience in leading teams in the analytics, measurement, research and database space to his role as Senior Director of Analytics at Olson. He utilizes analytical techniques to deliver data-driven insight and drive strategies for acquisition, retention, segmentation, targeting and other CRM tactics. Prior to joining the Olson team, Brad was Director, Analytics & Reporting for OptumHealth. He has also worked across a variety of industries including telecommunications, airline, automotive, railroad, healthcare, hospitality and retail for such clients as Amtrak, Wyndham Hotel Group, Mazda, AT&T and Electronic Arts. Brad resides in Plymouth, MN and has a B.A. in Mathematics from Saint Olaf College, and a M.S. in Statistics from Iowa State University.
Harnessing big data and machine learning to become more relevant in your customer touchpoints by identifying Share of Wallet opportunities in your customer base