Business Impact Modeling: A Closer Look

Business Impact Modeling is a family of statistical tools that can identify the effects of specific human capital practices and attributes on business performance. Although the statistical modeling in this approach is not very different from that used in ILM analysis, the sources of data are broader. Like ILM analysis, Business Impact Modeling makes use of data in HR information systems and payroll systems. However, it also uses data kept by the finance, quality control, marketing, and operations departments. Those data provide the measures of business performance.

The Production Function

What techniques constitute Business Impact Modeling? The following is a high-level overview of a few techniques. One fundamental technique is based on the production function, a core construct in microeconomics. The production function is a mathematical expression of the relationship between inputs and outputs in the production process. In its classic form, the left-hand side of the equation (the “outcome” or “dependent variable” side) is represented by the quantity of output produced during a particular period of time. This also may be represented in financial terms as value added: the difference between net revenues and material input costs. On the right-hand side (the “predictor” or “independent variable” side) appear measures of capital and labor.

In traditional microeconomic analyses, the right-hand side of the equation has represented labor (that is, human capital) too simply, all too often using a single measure such as headcount, hours worked, or compensation expense. For the production function to become a useful management tool, those simple measures have to be replaced with a greater number of specific measures of workforce attributes and management practices that influence labor productivity, precisely the type of information captured through an ILM analysis. Indeed, there is a long list of human capital attributes (length of service, diversity, education, prior work experience, etc.) and practices (hiring, base pay, variable pay, training, rotation through job assignments, etc.) that can be influenced or fully controlled by management, making the results of the analysis useful in setting strategic priorities. The relative weight of these factors can then be estimated through multivariate regression analysis. More detail on this approach is provided in Appendix B.

We adopted the production function approach to help determine the effects of workforce attributes and human capital practices at HealthCo, the hospital system discussed in Chapter 3. Using eight years of employee and performance data, covering 20 hospitals and facilities in their system, we were able to apply the production function to identify the key human capital drivers of workforce productivity, after accounting for differences in the capitalization of the hospitals and the patient mix they served. The analysis showed that over 60% of the differences in workforce productivity (measured as value added per employee) were explained by factors relating to human capital. Of these, the single biggest factor was how the facilities staffed their operations, particularly the mix of full-time versus part-time employees and use of overtime. You may recall Figure 3-2, the scatterplot diagram relating workforce productivity to the percentage of full-time employees in the workforce. This is a classic output of a statistical analysis based on the production function. In this instance it revealed that, all else being equal, the facilities that relied more heavily on part-timers experienced significantly lower productivity than those with a larger full-time staff. In its efforts to reduce expenses, the organization had overshot the optimal level of part-time staffing, because it failed to account for productivity in its cost calculations. The disciplined comparison of productivity within their system, made possible through this form of business impact modeling, exploded some myths about the drivers of labor cost at HealthCo that had serious policy repercussions.

Granger Causality

Another member of the Business Impact Modeling family of tools is Granger causality analysis. This method is especially powerful when one is focusing on the business impact of one or a few human capital issues. Employee turnover is an excellent example.

Most organizations blithely assume that turnover is bad. That untested assumption usually is expressed in a spreadsheet that attaches a dollar cost to each activity associated with employee turnover: time spent processing terminations, interviewing and screening applicants, adding new people to the payroll, and so on. Direct costs such as advertising and recruiters’ fees also are included. Those estimates are tallied and multiplied by the number of instances of employee turnover, resulting in a statement such as “Employee turnover cost this company $16,090,423.31 last year.” Note that this exercise is guaranteed to come up with a dollar figure, although that figure may contain lots of guesswork. This is also an exercise that implicitly assumes that (1) all turnover is costly and (2) all turnover affects the business negatively.

Quest Diagnostics is a company that did not settle for the usual assumptions about turnover. It wanted to know if turnover really mattered and if so, to what degree. Did it truly affect important business outcomes, or was it just a normal, if annoying, part of business? To get the answer the company undertook a Granger causality analysis of the impact of turnover on a number of measures of business performance, including quarterly operating margins for the company’s laboratory facilities.

The Granger approach rests on the observation that the best predictor of an outcome variable, such as operating margin, is the same variable in earlier time periods. Thus the operating margin in a current quarter should be relatively well predicted by the ongoing rate of prior quarters’ operating margins. Mathematically, this means that the right hand side of the model (the predictor side) also contains the operating margin. In evaluating the cost of turnover, the real test for this technique is to ask whether turnover in a prior period contributes significantly to the accuracy of predicting the current operating margin. If adding turnover to the right-hand side of the equation indeed increases the accuracy of prediction, one can be confident that turnover has a business impact. If that impact is large, turnover is clearly a problem that merits attention. Figure 6-1 illustrates the results for Quest Diagnostics.

Figure 6-1: Turnover Impact on Quest Diagnostics

Turnover was indeed a business problem. Further, the value of reducing it became known. More specifically, the financial return of actions and investments required to reduce turnover could be estimated because the model showed how much operating margin (and other outcomes) would improve for every percentage point reduction in turnover.

Myth Busters: Employee Turnover Is Not Always a Cost

Contrary to what most people believe, turnover can have many positive consequences. It provides an escape hatch for poor performers and creates vacancies for new and possibly outstanding employees. Turnover is also a means for improving the match between employees and their jobs and organizations. Without turnover, many workers would be stuck in jobs for which they are ill suited and companies would be saddled with employees who are not positioned to contribute maximally. Turnover also creates opportunities for advancement, especially when an organization is slow-growing. Indeed, it is a linchpin of career planning; companies in no-growth situations need turnover to open slots for high performers who are ready to advance.

Perhaps the most positive effect of turnover concerns the nature of human capital. Turnover allows companies to retool and replenish their workforces and thus to prevent stagnation. New ideas, new capabilities, and new experience from the outside generally are required if an organization hopes to remain competitive.

Any method used to estimate the costs of turnover has to account for potential positive effects like these, which traditional calculations do not do. It is far more useful to ask, “Is turnover really affecting the business?” and “Is there an optimal level of turnover for the company?” than to think that all turnover is bad.

The next step in solving the problem required identifying the drivers of turnover; this task was accomplished through an ILM analysis. All in all, a powerful business case for change was created. According to a report in People Management magazine, the company was able to reduce turnover and realize returns that were quite close to forecasts.^[1]

Other techniques that fit under the label of Business Impact Modeling include various forms of structural equation modeling, a powerful approach for establishing the direction and magnitude of effects over time, as well as formal evaluations of the business impact of selected management programs. Many of those evaluations are in effect evaluations of organizations’ experiments with new practices or programs.

National City Corporation, for example, evaluated the impact of a new training and on-boarding program designed to enhance workforce skills and customer service to support its new business strategy. The investment in the program was substantial. The evaluation of the program involved several components, including how much trainees learned and changes in their on-the-job behavior. The most compelling element was an assessment of the program’s impact on measures of business performance. Using a number of statistical modeling techniques, the company determined that the program raised business performance substantially, particularly the growth in the number of new checking and savings accounts and sales of annuity products. National City’s approach earned it the Corporate University Best-in-Class Award as well as the Optimas Award for Financial Impact given by Workforce magazine.

^[1]Cathy Cooper, “In for the Count,” People Management, October 12, 2000, 28–34.1.