Not All Analyses Are Good Analyses
As an engineering student, I participated in the cooperative education program, with many of my work semesters taking place at the NASA Research Center in Cleveland. It was there I learned the importance of effective data management and in collecting data from a scientific point of view. Maybe it was the fact that engineers spent a lot of their time setting up FORTRAN programs to do any kind of data analysis, but I learned to think through what I was looking for, because data analysis was expensive and time consuming.
Shortly after starting full-time employment in manufacturing, I bought an Apple II and learned VisiCalc. Then when my workplace started buying PCs, I was ready to use Lotus 1-2-3. By the time we were able to download information from the corporate systems into some spreadsheet-readable format, I was ready to work the data.
Average tenure? No problem. Forecast vacation load by department over the next three years? That will take just a minute, compared to the hours it used to take the payroll clerk. You want to see that in a graph? Give me 30 minutes and you will have what you need.
Today, of course, we can do so much more. But is there such a thing as too much data and too many tools to look at it? There are so many pieces of stored information about each employee that we can look at more combinations and permutations of data than we have time for.
The danger, I fear, is that managers are asking for data that is meant to justify decisions that they have already made. They ask for specific data to support an idea. And many of the people who can provide them that data are not in any way trained to understand the importance of objectivity in data analysis.
We are using data to tell the story we want to tell, not to investigate what the story is. By the time the data becomes a PowerPoint chart, the damage may be done.
We have amazing tools to analyze and understand data, as leaders we also have the accountability to assure that analysis is done correctly. Here are three things that a leader can look for the next time they are faced with a data-supported story.
- Overgeneralizations and one-data-point conclusions. “62% of our employees say their training is adequate for their job. We need to fix our training.” Um, maybe.
- Correlation is not causation. “The poorest performing teams work in Mike’s area. We need to get Mike a leadership coach.” Or perhaps we need to understand the overall working conditions in Mike’s area, including his approach to choosing supervisors and physical environment of Mike’s equipment.
- Data Dredging to find a hypothesis, and not to prove it.
The tools for reviewing and analyzing data are in nearly everyone’s hands. Being trained to use the tools, to create great pivot tables for example, is not the same as having the capability to keep the highest level of objectivity in interpreting data. People are not data points. Each chart we create tells us just one thing about a very complex system. Make sure you aren’t isolating data to tell the story you want to tell.
An engineer by training, Tim Gardner worked his way in HR when he realized the real optimization of manufacturing processes came from how the employees were managed. Currently the Director of Organizational Effectiveness for Kimberly-Clark Corporation, Tim has spent the majority of his career working with the creation and improvement of work teams, both hourly and professional. Tim is a member of the HR Bloggers Network, and his blog, The HR Introvert, looks at his HR work from the perspective of a non-traditional HR Pro. You can connect with him via LinkedIn and Twitter.