Blogs

Forecasting best practices: 3 approaches for financial executives

Financial Services Writer

When it comes to today's forecasting best practices, professionals have learned to apply extreme caution when charting simple correlations, which can be misleading. Financial executives need better techniques for strategic planning, forecasting and navigating volatile markets.

So how can financial leaders make better predictions? Three steps stand out:

1. Using probability-based modeling

The best proponent of this approach is Nate Silver, author of The Signal and The Noise: Why So Many Predictions Fail – But Some Don't, who brought Bayesian probability-based modeling out of the shadows and into executive offices. Financial executives increasingly use analytics to examine historical frequencies, which a Bayesian would call "priors," to assign a new set of probabilities to future events as new data arrives, rather than to say what the market will do next.

Beginning in the 1980s, this Bayesian forecasting received a boost with Black-Scholes options pricing models, which harnessed the computing power of advanced analytics to known mathematical techniques. Bayesian probabilities allow financial executives to continuously update their forward-looking models in real time. It's not a matter of, "This is what will happen," but rather, "Here is how our probabilities shift as events unfold."

https://kapost-files-prod.s3.amazonaws.com/uploads/direct/1444753382-43-3431/ForecastingBestPractices_Blog.jpg2. Leveraging robust statistical back-testing

In order to establish the best Bayesian priors, of course, financial executives need to know which factors to focus on and which new pieces of data will affect outcomes. That's best done through robust statistical back-testing.

As Oliver Churchill, cofounder and CEO of Acuity Sales Decision Science, writes on LinkedIn, it's too easy to either tell outright lies or, worse, to convince ourselves of things that aren't completely true when a simple chart appears to support our view.

Churchill notes that forecasting best practices always involve multi-dimensional review and modeling of data. We, being humans with inherent biases, cannot accurately distinguish the effects of certain factors on financial success. As a result, poor decisions tend to be made according to the relative political heft of the data-reviewer within the organization or according to the clash of human incentives. Think bonuses, promotions or the boss's favorite project.

Statistical modeling done right, however, doesn't confirm our preexisting biases, but rather provides an impartial review of past influences on outcomes. Despite these benefits, not many companies actually use back-tested statistical modeling to understand their sales inputs while closing a deal. Financial executives need to incorporate this approach to ensure they make the best decisions going forward.

3. Remaining open to nontraditional data

In other situations, the existing data sets regarding a financial decision prove insufficient for our needs. In this case, companies need to build their own databases, which can in turn become a source of proprietary strength.

Temo Maldonado, lead underwriter at LiftFund, the largest nonprofit business microlender in the country, explained in a recent interview with me that traditional underwriting data inputs prove insufficient for their purposes. Liftfund's mission of providing business loans from for-profit banks to unqualified customers meant they had to seek alternate data.

The microlender found that using nontraditional inputs such as alternative collateral, outside income sources and social networks, combined with traditional underwriting data, could drive their Microloan Management Services (MMS), a custom-built loan application screening tool. Maldonado credits the rigorous analytics of MMS, which incorporates all these inputs, as the key to their growth and ability to make MMS available to 13 other microlenders nationwide.

Financial leaders will do well to incorporate nontraditional data with option-based modeling and statistical back-testing in order to mitigate market volatility and make better forward-looking decisions.

Explore IBM Financial Performance Management solutions and learn how you can transform your organization's planning, budgeting, forecasting, reporting and analysis processes into more dynamic, efficient and connected experiences.