Moving from possibility to probability with analytics


Share on LinkedIn

What could possibly happen to your organization performance results? At the operational level sales order volume could be up or down. Prices of purchased commodity materials like steel or coffee could be up or down. On a strategic macroeconomic level, consumer demand could be up or down. From a risk management perspective, weather fluctuations could adversely affect the best laid plans.

How could you know the impact, including the financial impact, as these possibilities occur at various levels? There are three broad ways: a single best guess; the worst, baseline, and best likely outcomes; and a probabilistic scenario of the full range of outcomes. They all include predictions with analytics.

(1) Single best guess

Most organization plan for results based on their manager’s best assumptions of what they estimate. For example, in the annual budgeting exercise managers forecast sales mix volume, labor rates, and prices of purchases. Each is a single point estimate, and the accountants aggregate them to produce a single budget.

(2) Worst, baseline, and best likely outcomes

The more advanced organizations consider three ranges of outcomes: worst, baseline, and best likely outcomes. Separate predictions for made for the key variables in the plan. Then the three overall possibilities are calculated. This provides a sense of the range of outcomes. These organizations might individually test the sensitivity of the key variables by increasing or decreasing one of them – one at a time.

(3) Multiple probabilistic scenarios

The most advanced organizations take this process to its ultimate limit: from three scenarios to the full range of possibilities. That is, they estimate the probability distribution of each variable, perhaps as percentage increments from the base (e.g., -20%, -10, 0% base, +10%, +20%). By combining these, they move from the three single point outcomes to viewing a distribution curve of dozens and conceivably hundreds or thousands of outcomes. The benefit is they can have more certainty of the increasingly uncertain world they operate in. In addition, the variables become understood as “drivers” of the results where the level of each one may be able to be proactively managed in advance of their occurrence.

Imagining the future

The breadth and granularity of the distribution curve increases as the probability ranges for each variable is segmented, as more variables (not just the key ones) are added, and as each variable is sub-divided (e.g., from a product family to its individual products). The three scenario approach gives a limited view of risk in contrast to the multiple probabilistic distribution curve. With the latter, sensitivity analysis can become very refined, including automated increases and decreases of each variable to determine which variable drivers are more impacting.

Now take this process to an even higher level by increasing the time interval frequency of re-forecasting one or more (or even all) of the variable drivers.

What influences the accuracy and quality of the distribution curve? A critical one is the forecasting of each variable. If the baseline is way off then incrementing it up or down is also going to include error.

To achieve this “best practices” approach requires a combination of advanced analytics, reliable forecasting techniques (e.g., monte carlo methods), and a powerful computational software engine. If this is supplemented with robust reporting, visualization, and analytical power then it is nirvana. The full range of probabilistic outcomes can be viewed and at more frequent time intervals approaching near real time. The benefits are endless. Risk management becomes scientific. Rolling financial forecasts replace static and fixed-in-time annual budgets the quickly become obsolete. Drivers can be proactively managed such as supply chain logistics and inventory management.

Predictive analytics is becoming a “hot” term with enterprise performance management. With this opportunity to move from just discussing the possibilities to understanding the factors impacting your organization and also taking actions based on the interdependent probabilities, are you surprised? You shift from possibility to probability – managed probability – of outcomes.

Republished with author's permission from original post.

Gary Cokins, CPIM
Gary Cokins (Cornell University BS IE/OR, 1971; Northwestern University Kellogg MBA 1974) is an internationally recognized expert, speaker, and author in advanced cost management and enterprise performance and risk management (EPM/ERM) systems. He is the founder of Analytics-Based Performance Management LLC, an advisory firm located in Cary, North Carolina at Gary is the Executive in Residence of the Institute of Management Accountants (


Please use comments to add value to the discussion. Maximum one link to an educational blog post or article. We will NOT PUBLISH brief comments like "good post," comments that mainly promote links, or comments with links to companies, products, or services.

Please enter your comment!
Please enter your name here