We’re always forecasting — thinking about what will happen, assessing its likelihood, and contemplating the implications. For CFOs, the stakes are especially high when it comes to the difference between accurate and inaccurate forecasts.
It’s a CFO’s worst nightmare: Expected and actual results fail to converge, and managers are left unable to understand the cause and adapt to the variance. Numerous “why” questions are asked, but often there are no good answers. Is there a fundamental problem with the appeal of our products/services? Did a competitor make a move we didn’t expect? Is it due to macro factors that are out of our control? How much does it matter to the next forecast on the future value of our business? It can be difficult to find viable signals amid the noise.
Fortunately, major advances in data collection and computing power have facilitated the use of analytical techniques that enable companies to better capture leading indicators, understand the correlation and causality of factors affecting business performance, and generally produce more precise forecasts. These new tools and capabilities should allow CFOs to better answer the “why” questions and take appropriate action.
Thinking about it in simple terms, there are three key prerequisites for adequately reimagining forecasting.
The biggest challenge for many CFOs is aligning the C-suite on the best drivers from each business area to inform forecasts. From there, consensus will still be required on how to analyze results and quickly leverage those insights to drive strategic or operational adjustments across multiple areas.
It’s crucial for organizations to establish a framework for translating forecasts into actionable insight and ultimately to action — what we call the “mind-muscle connection.” To do this, the forecaster must prioritize the data to be explored and analytical techniques to be applied based on what generates that most timely and actionable results for senior decision makers.
It’s also critical to cultivate an organization mindset that minimizes bias as well as values quality, consistency, and accuracy. For example, mid-to-long-range forecasts based on alternative market, competition, and strategy scenarios benefit from using samples of data from more sources, as well as analytic methods that allow hundreds of simulations to be run, evaluated, and re-run based on the collective judgement, experience, and data-driven insights from the C-suite.
The search for better forecast indicators may feel to some CFOs like wandering through endless data wastelands that don’t provide sufficient bang for the buck. As a result, many forecasts are built on relatively thin, high-level historical time-series using inefficient or in some cases improper modeling techniques. Yet it’s now often possible — and increasingly crucial — to build more real-time signal data into models in order to provide more insightful forecasts, increase precision, and understand day-to-day sensitivities.
For example, companies are now using web-search and weather data as leading indicators of economic activity and demand — sometimes even at a zip-code or retail-store level. Others are using social networks of individuals with smartphones to capture granular consumer behavior data so as to better inform marketing messages and pricing offers. Geo-satellite and drone imagery is being used to inform catastrophe planning. Sensor and unstructured work log data are being used in asset intensive industries to detect anomalies, avoid operations disruptions, and inform recalls. Image data is being used to understand customer reactions to new product designs.
CFOs should use “a test and learn” approach to seek out and systematically evaluate the particular data that provides the leading indicators of their organization’s performance.
With the right data in hand, a range of forecasting techniques can be used and at times combined for the purpose of creating better forecasts that become more predictive, insightful, and in some cases prescriptive. The key is matching and combining the appropriate techniques based on the type of data available, the desired accuracy, and the cost.
A CFO wondering about the impact of a rise in interest rates on deposit pricing may leverage combinations of time series composition, hierarchical forecasting, and agent-based and system dynamics models to understand the impact of long- and short-term interest rate changes on the cost of funds and implications for deposit pricing strategies. Taking advantage of these techniques will require CFOs to expand their team’s analytics capabilities, judgment and sophistication in areas of data science, statistical modeling and computation mathematics.
In addition to the three prerequisites discussed above, the right organizational mindset is required in order to build greater accountability around forecasts and responsiveness to results. Engagement and collaboration across various groups are critical to establishing accurate forecasts based on an unbiased consensus outlook. Learning from failure, as opposed to persecuting it, is key. Finally, adopting an “In God we trust, all others must bring data, an open mind, and judgment” mentality is needed to answer “why questions” that are at the heart of improving performance. A good place to start with the mindset shift is in the C-suite.
Reimagining forecasting may seem like a gratuitous exercise in crystal ball design. But it is becoming an important step for CFOs to take, in order to make sure their companies are using leading indicators to maintain their positions as market-leading companies, strive to innovate, and protect the brand — as opposed to lagging the pack with lagging indicators.
Paul Blase is data and analytics leader, Global and US Consulting, at PricewaterhouseCoopers.