By Steven T. Hunt, Ph.D.
Peter Howes, a pioneer in HR data analytics once told me that a little data in the wrong context is often worse than no data at all. If people misinterpret data, it strengthens their beliefs in the accuracy of something that is not true. This is a risk when sharing HR data with business leaders. Consider the following story. Following a financial downturn, a large company had to rapidly reduce total workforce costs. Senior leaders were given spreadsheets showing salary and headcount for different departments. They identified a team working on a new but non-critical product that had relatively high workforce costs. Eliminating this team was an easy way to remove costs based on HR data showing job titles, functions and salaries. But the leaders never looked at data showing the capabilities of the people on the team. Shortly after letting the team go, the company discovered it had eliminated several highly skilled engineers. These employees were on this team because of their high performance and creativity. Several of the employees possessed unique skills critical to the company’s business operations. A few months later the company had to re-hire these employees as consultants at rates far greater than what they were paid as full-time employees. And their sense of company commitment had been lost.
It might look like the leaders in this story made a stupid decision. But these leaders were not stupid people. They were smart people who made a confident decision based on accurate data interpreted the wrong way. What they lacked was additional data needed to fully understand the context of their decision and its impact, both positive and negative. Part of the art of using HR data is presenting it in a way that leads people to draw appropriate insights and conclusions. This is about providing data in the right context coupled with effective analytical interpretation. To illustrate this concept, consider the following four stories that demonstrate ineffective to effective use of HR data.
Presenting HR data without any business context. Jenny Dearborn, author of the book “Data Driven” told me a great story that illustrates the risk of presenting HR data without any business context. Early in her career, she shared data with a business leader indicating how many people in his organization had completed various training programs. Rather than responding positively, the leader said, “all this data tells me for sure is that dozens of people in my company have spent hundreds of hours sitting in training classes instead of focusing on getting actual work done.” The leader did not say that training was a waste of time. But he made it clear that this HR data was not valuable unless it could be linked to business-relevant data such as sales productivity.
Presenting HR data without enough business context. Hiring freezes are a common method used to control costs. From a financial perspective, this makes a lot sense. By stopping hiring, companies swiftly reduce growth in operating costs caused by salary. What leaders implementing hiring freezes do not see is the financial losses they create. This starts with wasting the months of time spent recruiting skilled candidates, only to tell them they cannot be hired. The best candidates usually have multiple offers, so when a company implements a hiring freeze, it is giving many of its top candidates to other companies. The company may never recover from this loss. Employees are hired because companies need people to generate revenue and run efficient operations. The revenue and efficiency that would be gained by hiring employees is delayed and potentially lost forever when a hiring freeze goes into effect. The problem with hiring freezes is not that they never make sense. The problem is they are usually implemented based solely on data showing the cost saved by eliminating new hires, without considering contextual data showing the financial gains that new hires generate.
Presenting HR data with the right amount of context. A major challenge for business operations is determining the right number of employees to hire to maximize profit and growth without generating excessive workforce costs. This is particularly true in low margin industries like retail where small differences in operating costs can make the difference between profit and loss. I once worked with a retail organization that made creative use of HR data to determine the optimal number of store managers to hire in a region. Historically the company strove to keep costs low by only hiring store managers when there was an existing vacancy within the region. This meant the company did not hire new store managers until an existing store manager quit. An intrepid HR leader observed that when a store manager quit, the morale of employees in that store often suffered, turnover increased, and sales declined. Pressure was also placed on adjacent store managers to cover the store until a replacement manager was hired, and consequently, the performance of adjacent stores suffered as well. He used store sales and turnover data to demonstrate that the cost incurred by waiting to hire store managers until after an existing manager quit was greater than the cost of employing an extra “floating” store manager who could immediately step in and run a store if the manager left. This story illustrates the benefits of looking beyond simple HR data like staffing and salary and incorporating contextual data that illustrates how people impact broader business operations.
Presenting HR data with the right amount of context plus analytical insights. Some HR data needs no explanation. Business leaders do not need much guidance to understand the impact of HR data demonstrating things such as “high performers are quitting at twice the rate of low performers.” As one statistician told me, the best data has inter-ocular significance – its meaning hits you right between the eyes! However, not all HR data lends itself to such easy interpretation. The last few years have seen tremendous growth in the use of machine learning and advanced non-linear mathematical modeling methods to draw analytical insights from HR datasets that might otherwise be overlooked.
Analytical methods like machine learning do not magically turn HR data into meaningful information. The data must meet certain conditions related to data quality and sample size. It must also be tied to business metrics and presented in the right context. But given these preconditions, advanced analytical methods can greatly enhance the value of HR data. For example, I once worked with a company to develop staffing assessment tools to predict sales performance in shoe stores. The company had observed that better salespeople tended to be socially confident (e.g., they initiate conversations) and personable (e.g., they are interested in learning about others). Based on this, the company favored hiring candidates who were socially confident and personable. After the company amassed data on several thousand hires, they used non-linear mathematical models to determine the optimal fit between candidate characteristics and sales. The analysis confirmed that good salespeople did tend to be socially confident and personable. But the absolute best salespeople were socially confident but not very personable! Most socially confident people tend also to be personable. But that rare unpersonable, socially confident individuals sold the most shoes because they were highly task focused. They were not striking up conversations with customers to learn about them; they were starting conversations to sell them shoes. This insight made total sense to the store managers once it was called out. But without the use of advanced analytics, the company never would have made this realization and modified its hiring profile.
A little HR data can be a dangerous thing. But it can also be a powerful force for good. The key lies not in the data itself, but in ensuring the data is placed in the right context and interpreted in the right way at the right time.