simulate performance improvements and then estimate the financial return of running rank and yank in a “call-center-like environment” over a 5-year period

QUESTION

what is the manipulate IV?

This study was designed to simulate performance improvements and then estimate the financial return of running rank and yank in a “call-center-like environment” over a 5-year period. We ran the simulation under two different scenarios. The first baseline simulation assumed no voluntary turnover. The second scenario added 30% voluntary turnover because voluntary turnover is a huge operating challenge for call centers and because it has the effect of reducing gains from any productivity intervention (Adsit & Pyzdek, 2008).

ANSWER

Simulation Variables

Productivity. Average productivity for the workforce as a whole was our primary dependent variable. We did not focus on a particular call-center measure like talk time or length of after-call work but rather on an overall assessment of the productivity of call-center agents and what that might be worth to an organization. We judged this approach to be realistic. Research has demonstrated that small composites of objective performance measures can have extremely high reliabilities in jobs similar to call centers (Crant, 1995). Based on Schmidt and Hunter (1983), we used 40% of worker salary as the value for one standard deviation improvement in worker productivity. Although there are other estimates available of the standard deviation of performance (see both Global Estimation and Cascio-Ramos Estimate of Performance in Dollars (CREPID) in Cascio, 1991), managers have been shown to be accepting of the 40% of salary estimate (Hazer & Highhouse, 1997).

Our model also factored in changes in worker performance over time. First, our financial model included the impact of lower performance from the inexperienced new hires during their first year. These estimates are part of the increased costs related to initial underperformance of work outlined by Hillmer et al. (2004). Additionally, we modeled year-to-year performance variability. This year-to-year performance instability covered average individual performance increases, decreases, and instabilities between years and was based on research data from call centers and other similar jobs (see the Performance Variability section, below).

Voluntary turnover rate. Any simulation attempting to model a call-center environment needs to include voluntary turnover as a key variable. There are two reasons for this. First, if you look at turnover across functional groups inside any given company, the turnover rate for the CSRs in the call centers is likely to be significantly higher. Wages are low and the work is repetitive, boring, and highly stressful (Adsit, 2013). The second reason is that past simulations have shown that voluntary turnover has a dramatic effect on performance and/or productivity gains: As turnover increases, performance drops and productivity improvements can quickly plateau (Adsit & Pyzdek, 2008; Scullen et al., 2005).

As an industry, call centers have a wide range of voluntary turnover influenced by many factors including location, management style, industry served, unionization, and so forth (Martí-Audí et al., 2013). The average rate of voluntary turnover across the industry for call centers in Australia was estimated at 29.5% in 2001, ranging up to 59% for particularly high-stress environments (Siong, Mellor, Moore, & Firth, 2006). Industry-wide estimates in 1999 were at 30% (Stuller, 1999), and 2005 estimates ranged from 40% to 70% (Baker, 2005).

We used 30% as our estimate of voluntary turnover for our second simulation to have a widely applicable and middle-of-the-road estimate. We assumed that voluntary turnover is randomly distributed, based on research that showed that voluntary turnover is not correlated with perfor- mance (Mulligan & Schaefer, 2011).

Percent fired by rank-and-yank system. Common firing rates for rank-and-yank systems also vary, but two widely used numbers are 5% and 10% per cycle (Scullen et al., 2005). We used the 10% originally proposed by Jack Welch, which is commonly used in rank-and-yank systems (Mulligan & Schaefer, 2011).

Voluntary-turnover cost. Estimates for costs of turnover irrespective of industry vary enormously. However, these estimates likely all depend on employee position within the company as well as industry itself. Therefore, we used a call-center-specific estimate of turnover costs. Hillmer et al. (2004) suggested that turnover cost for call centers would be 82% of a CSR’s salary for every voluntary departure.

Rank-and-yank turnover cost. The costs associated with rank-and-yank turnover are different than the costs associated with voluntary turnover. The chief differences with rank-and-yank turnover are that it happens all at once and employees know when it is coming. Voluntary turnover happens randomly and is spread out. A group of new recruits could easily be hired weeks earlier, trained, and be ready to get on the phones the Monday after a scheduled rank and yank occurred.

Thus, to quantify the rank-and-yank turnover costs, Hillmer et al.’s (2004) voluntary-turnover cost estimates needed to be modified. Much of Hillmer et al.’s cost of voluntary turnover related to the work not being completed by the now-absent employee for an average length of 2 months (until the hiring process for a replacement was completed). Specifically, the estimate included the direct cost of the vacancy, lost productivity from stress on remaining agents, adjustments to infrastructure, and reduced performance from the leaving CSR prior to departure. We judged that these specific direct and indirect costs would not apply to a planned rank-and-yank situation, and therefore we eliminated them from our calculations.

Hillmer et al. (2004) listed other voluntary-turnover costs that apply to employees fired by the rank-and-yank process. Specifically, the direct costs included screening potential candidates, interviewing them, testing them, training them, and orientating them. The less tangible costs included error rework, increased supervision and coaching, and the CSR’s learning curve over the first year. Based on this rational and using the Hillmer et al. (2004) estimates, we calculated the cost of a planned rank-and-yank turnover to be 46% of a CSR’s salary over 1 year.

You will note that our cost estimates do include HR costs in the form of recruiting, screening, training, and so forth. Some might argue that we have underestimated the costs of running a rank-and-yank program. We think our estimates our defensible because we are running this program in a call-center environment. Call centers are already constantly evaluating agent performance and they are primed and staffed for high turnover. Yes, handling an additional 10% turnover would add costs, but those additional costs have been accounted for in our estimates.

Performance Variability

In the performance literature, there is no question that work performance over time can be highly variable. In call centers, CSR performance is generally measured by counting behaviors and quantifiable outcomes that are directly tied to organization success. This can include such things as calls answered per hour, calls placed on hold, time spent on after-call work, and so forth. We were unable to find published studies of call-center long-term performance variability, so instead we used the extensive literature on objective measures of sales performance.

A meta-analysis of sales performance looked at the reliability of performance measurements over time and found that the longer the period and the simpler the task, the less the performance measurements varied (Hunter, Schmidt, & Judiesch, 1990). This meta-analysis generated a reliability of .89 for objective performance measures of general-sales jobs when measured over a year-long period. However, single measures may not tell the whole story in some cases. Composites of objective sales-performance indicators can also generate extremely high long-term reliabilities— three objective sales performance indicators showing a reliability of .90 when averaged over a 9-month period (Crant, 1995). This suggests that in these types of jobs, performance variability is very low when judged over long periods of time. But it also begs the question: Are these sales results good models for CSR performance variability?

One of the authors of this paper has access to internal studies from two different former call-center clients who have agreed to anonymously share performance variability outcomes. These unpublished CSR performance reliabilities seem to outperform the published estimates of sales reliability. In company A, customer-reported satisfaction over 2 years was .91 (n=133), first-call resolution over 2 years was .92 (n= 106), and average handle time over 2 years was .98 (n=105). In company B, task resolution over 4 months was .98 (n= 66). This suggests low performance variability in call centers over long periods of time, especially for objective measures. However, even though the reliabilities for call centers seem to be higher than sales, we used a performance reliability of .85 in our simulation to have a more conservative and defensible estimate.