Demand forecasting—the ability to project demand by role and project—is an important part of resource management. Companies conduct demand forecasting in several ways. Some have project managers enter demand by role for, say, 12 months. Others communicate standards to help drive consistency in how project managers estimate demand. Still, others create demand forecasting algorithms to help drive greater accuracy. An algorithm is an equation that handles variables. There are many different types of algorithms in terms of the equations and the lookup tables they reference.
Obstacles to Overcome
For all algorithm structures, companies are likely to encounter one or more of these obstacles:
- First, algorithms are data intensive. They require higher-quality data, clear data standards, authoritative sources of data, and just general data management.
- Second, algorithms create a “black box” feeling with the business. Often, the business doesn’t understand algorithms. They view the process as “Data goes in. The system spits out a number. And I don’t know or understand how it was derived.” This insight holds true even when stakeholders have provided requirements and direction around the algorithm structure.
- Third, algorithms require regular tweaking to improve them. That can be costly.
. . . algorithms won’t deliver accuracy higher than 90%. No matter what anyone tells you, it’s just not going to happen.
It’s true that algorithms can provide more accurate estimates. However, the question remains: How accurate do you really need to be? Manually generated estimates are often neglected. This naturally leads to inaccuracy and diminished effectiveness. But even algorithms won’t deliver accuracy higher than 90%. No matter what anyone tells you, it’s just not going to happen.
Do Algorithms Have a Place in Effective Resource Management
All of that isn’t to say that algorithms have no place in effective resource management. You can house algorithms at different levels of the work breakdown structure in your timelines. You can track actual effort for time tracking on one level and perform resource forecasting at another level.
In consulting with companies, I frequently recommend a walk-before-you-run approach. Apply some standardization to manual estimates before you adopt algorithms. As time moves on, try a pilot with algorithms and gauge its value relative to the effort to maintain and explain the algorithm and its output.