Guiding Principles to Build a Demand Forecast
Demand forecasting is key for many industries, including finance, healthcare, and retails, and it is one of the most challenging tasks for predictive analytics. We review challenges and guiding principles of demand forecasting.
Demand forecasting is one of the most challenging fields of predictive analytics. This is a reality that is industry agnostic – true across finance, health care, and consumer goods and retail. This will not come as a surprise to business decision makers and data scientists working hard to leverage that information.
While each industry has its own challenges – retail demand may be far more sensitive to economic pressures, for example, while health care is adapting to different delivery models and regulatory requirements – there are common pitfalls. Fortunately, there are common strategies to optimize the benefits of demand forecasting.
The first challenge is complexity, an issue that is often poorly understood across the entire enterprise. Most businesses have enough logistical, transactional and other data, although in some cases this information is still not as well-developed – or meaningfully communicated across units – with a strategic view toward the necessary forecasting endpoints. In today’s data-driven environment, smart business leaders have moved beyond the basics of time-series or other historic rates in the hopes of applying the information toward robust future growth opportunities. But even good data models exist in an evolving environment of consumer behavior shifts, cyclical economic fluctuations and more factors that make it difficult to identify critical trends. Predictive analytics products are available to try to solve for the issue.
What these products cannot solve as readily are the internal dynamics that characterize a business – especially where models have reflected a lack of consistency in the data, because the assumptions and drivers of one department or operational unit are not aligned with those of the others. Whether that’s a function of the organizational culture and a politically-charged environment, a failure of legitimate but conflicting visions, or poorly communicated agendas and goals, the resulting forecast mismatch can be expensive.
Think for a moment about the kinds of questions you ask yourself about forecasting. When you make demand decisions, are they based on the same evaluation drivers that are used in other decision-making units? Do they include the level of complexity that relevant data from those units would offer? The first rule for optimizing your demand forecasting capacities is to be consistent. A uniform approach to what data you assess and what it’s for, applied across all categories, creates a foundation for the “one truth” solution on which you can reliably and confidently base your future-demand strategies and resources.
Developing a model that’s consistent, comprehensive, clear and driven by the relevant causal factors doesn’t just happen once, because the success driven by the high-quality demand forecasting that you seek often requires a more iterative process. But it doesn’t need to be intimidating, either. Many businesses seeking to capitalize on the value of “big data” continue to refine their understanding of how that’s defined – and far more, in how predictive analytics are applied in order to drive growth. It’s more important to have leaders across the enterprise understand that the demand-modeling designs are likely to be more fluid, more responsive to actual use and application, and therefore adjust in time.
But therein lies the real success story of demand forecasting: an organization that really uses it, and reaps the benefits. Algorithms designed to maximize your profitability and value will not work until you have created the culture that welcomes them. In other words, that requires buy-in at all levels and across all units of your business. Leaders need to prioritize how they are integrating their demand forecasting strategy into the existing framework of the skeptical and budget-wary heads in sales, marketing, transportation, and perhaps especially HR and training – since the real investment in demand models and the objective data they provide is in their acceptance by the people who use them.
So what are the five main guiding principles to creating a robust demand forecasting solution?
- Objectivity – Forecasting models should serve as a tool to get to an objective and an unbiased number that can be defended with concrete data
- Consistent Approach – A similar approach to forecasting should be used across brands, markets and categories; whenever possible, similar processes should be used by different functions
- Transparent Assumptions – Key assumptions on market and execution drivers (economy, inflation, price, distribution, marketing, account growth, etc.)
- Business User Engagement – Creating buy-in of business users is critical to successful implementation of forecasting analytics
- Embedded Processes – Forecast analytics need be embedded in regular forecasting processes
- Amazon: Research Scientist, Forecasting
- Kaggle Epilepsy Seizure Prediction Challenge
- ICON Challenge on Forecasting and Scheduling