« View all press releases

The Darwinian Workplace.


by Serguei Netessine and Valery Yakubovich - Harvard Business Review – April 18, 2012

Dale Carnegie told a story about the steel magnate Charles M. Schwab. One evening, to incentivize the workers in a mill, Schwab wrote on the floor the amount of steel the day shift had produced. Seeing the number, the night shift worked hard to top it, marking its own figure down. Soon the two shifts were vying for bragging rights, and production soared. “The way to get things done,” Schwab said, “is to stimulate competition.”

Servers at the Massachusetts-based restaurant chain Not Your Average Joe’s don’t find numbers chalked on the floor, but they always know how they’re doing relative to their colleagues, owing to a cutting-edge workforce management system. Rather than forecasting demand and staffing a restaurant accordingly, as most systems do, the software tracks waitstaff performance in terms of per-customer sales and satisfaction (gauged by tips). Highly rated servers are given more tables and preferred schedules. By shifting work to its best servers, the restaurant hopes to increase profits and motivate all employees.

The View from One CEO—and One Server

Not Your Average Joe’s is a chain of 17 casual restaurants, most of them near Boston.

CEO Stephen Silverstein founded the firm in 1994.

HBR: What was your goal in introducing this system?

Silverstein: Servers are our sales force. We thought it would provide insight into an area where we had no data and would drive sales and profitability.

What was the reaction among staff members?

It wasn’t as negative as I expected. Many loved seeing their results. This is a helpful tool for servers to increase their sales and tips. After all, servers are commissioned salespeople.

Has it caused turnover?

Very few employees have quit. Some low-performing servers will be asked to leave if we can’t help them improve, but losing the weak is good for the company.

What’s the financial impact?

It’s too early to tell for sure, but we think we will raise check averages by 2% to 3%, from $17 to $17.50 per guest. Based on 60,000 transactions a week, that adds up to about $1.5 million a year; with a 40% margin, the return is tremendous. We’ve also seen that guests prefer the higher-scoring servers, and since the system should help all servers improve, we hope that guests will be returning more frequently.

Colleen Cushman is a food server in Beverly, Massachusetts.

HBR: How did you rank in your first few shifts?

Cushman: Initially I ranked maybe 15th out of 30 servers. I’d never waitressed before, so my first concern was not dropping anything. But over time I learned to sell, and my rank went up. Instead of just having a customer look over the menu, I’d sell specific items: “This appetizer is my personal favorite.” I like suggestive selling. I’m very indecisive, so when I’m a diner and I hear servers describe different options, it helps me decide.

How do you benefit from the rankings?

There are usually 12 servers per shift, so as long as I rank in the top 12, I’ll get the shifts I want. Fridays, Saturdays, and Sundays are the moneymaking days, when I want to work.

How do low performers react to the rankings?

They often aren’t even aware of the rankings—they’re typically not on top of that.

Does your ranking ever fall?

Sometimes. I might get a party of 12 teenagers who want separate checks. Average check size is a factor in the rankings, so that can bring you down.

Are coworkers competitive about the rankings?

We joke back and forth, but it’s nothing intense. For the most part, this system helps everyone get better, and a little friendly competition never hurts.

Labor is one of the biggest expenses in industries such as restaurants, retail, and call centers—but it has proved difficult to get workers to increase their productivity as they gain experience. For decades managers have paid these workers as little as possible and accepted the high turnover that ensued. In our research and consulting, we’re seeing new human resources practices that take a systematic, quantitative approach to raising employee productivity. By using technology to create a form of the leaderboard typical in sales organizations, innovative firms are infusing their workplaces with competitive spirit. Both companies and high-performing employees stand to gain. We call these firms “winners take all” organizations.

Instead of distributing work evenly among employees, winners-take-all organizations allocate according to merit: Better workers take more assignments, and the others get what remains. The model exploits the fact that workers differ dramatically in productivity because of such factors as skills and attitude, which can be hard to assess when hiring. Over time, it may induce low performers to quit, leading to a higher-performing workforce and a constantly rising bar.

Shifting work in this way can significantly boost revenue. Let’s imagine two waitresses, Alison and Dana. Alison brings in $10 more per check, on average, than other servers, and she receives 4% more in tips (on equivalent checks). Dana’s checks are $20 less, on average, than those of other servers, and her tips lag by 4%. The classic HR approach would be to roll out a training program for all servers. A winners-take-all organization would reallocate work from Dana to Alison. For every table shifted, it would take in, on average, $30 more—the amount by which Alison outsells Dana.

In fields such as sports, fashion, and education, ranking talent is nothing new. It may seem far-fetched for industries populated by low-wage workers, but in our experience it works well there, too: Underperforming workers often have the skills to do better and simply lack motivation. And because ranking systems can bolster company revenue, all workers can win in the long run.

Not every industry is suited to a winners-take-all approach. It’s necessary to be able to measure each worker’s performance accurately and to have a system with few interdependencies among workers’ efforts. We’d recommend the approach for car dealerships and hair salons, for example, but not for research institutions.

And it’s important to design the system carefully. Consider what happened at one well-known retail clothing chain. For three years its stores have used a proprietary system to schedule the most productive sellers—in terms of average sales per hour, units sold, and dollars per transaction—during the busiest times. This rewarded the best employees with more hours (most employees are part-time) and hence more pay and led to higher sales rates for all employees. However, it engendered anxiety and excessive competition, and the company had to adjust its system as a result (see the sidebar “Getting the System Right”).

Getting the System Right

When one clothing retailer shifted to a productivity-based system, it faced a host of obstacles, ranging from conflict among employees to confusion over incentives. Here’s what we’d recommend to solve the problems that arose:

Excessive Competition
Challenge: Giving top sellers preferred schedules can create an overly aggressive staff whose members try to steal clients from one another.

Solution: Clients must be divided up fairly. Use a set rotation to determine the order in which sales associates serve walk-in customers.

Unpredictable Swings in Pay
Challenge: Employees may see their pay vary widely and unpredictably when performance drives the number of hours each person works.

Solution: Average employees’ performance over large periods of time. For example, base assessments on a week of work, not a single shift.

Unfair Comparisons
Challenge: Workers unable to take peak shifts are at a disadvantage, because it’s impossible to achieve comparable sales during off-hours.

Solution: Performance calculations must be adjusted for the period of time worked and should control for the predictable sales lift during busy hours.

Erratic Work Schedules
Challenge: Employees may not like or be able to accommodate schedules that are constantly changing according to performance.

Solution: Systems must take into account employee preferences and constraints, including each employee’s “dream schedule.”

What are the key challenges when converting to a winners-take-all model? On the basis of our extensive involvement with some early adopters, we recommend considering the following:

Some settings are inherently more appropriate than others for winners-take-all transformations. One of the organizations we studied was the virtual call center LiveOps. Its agents interact with customers about a range of products with varying conversion rates. For items such as insurance policies and complex dietary supplements, which have low conversion rates, shifting calls from the bottom 10% of agents to the top 10% lifted revenue by as much as 18%. But for high-conversion products, such as fitness training programs, such reallocations had less payoff, because most callers placed an order no matter who answered the phone. The phenomenon holds for other businesses as well: If the likelihood and size of sales don’t vary according to who’s servicing the client, there will be little benefit from a winners-take-all system.

Workers need specific information to help them understand how to perform better and what benefits might accrue. The software used at Not Your Average Joe’s, created by the start-up company Objective Logistics (in which Serguei Netessine, a coauthor of this piece, owns an equity stake of less than 1%), conveys explicit data: It predicts that if a waiter sells, say, nine martinis, 10 appetizers, and 50 salads during a Wednesday-night shift, he will earn $125 more in tips and rank first among the shift’s servers.

Workers in traditional organizations transitioning to a winners-take-all model may complain that the new system “dehumanizes the managerial process.” In our experience, this problem is far less common in nontraditional settings. The agents for LiveOps, for instance, are independent contractors. They’re geographically dispersed, work separately from one another, and have an arm’s-length, computer-mediated relationship with the company. They know their own positions on the performance curve but are not aware of other agents’ individual rankings. Under these conditions, competition is unlikely to turn into destructive interpersonal rivalry.

Performance naturally fluctuates over time. To avoid unfairly penalizing employees, managers should beware of reading too much into the short term. The Objective Logistics software tracks waiter performance over several weeks; LiveOps assesses agent performance over 25 to 75 calls for each product.

It’s relative performance that counts. Workers need to realize that although their performance may be consistent, their ranking, and therefore their workload, can change depending on how other employees do. And firms must recognize that not all tasks can be tracked in exactly the same way. A restaurant may judge performance during a breakfast shift differently from performance during a dinner shift. A call center may vary its assessments according to the products being sold.

New hires need special accommodation. If a worker starts at the bottom, he may take a long time to move up—and may quit before he can acquire firm-specific knowledge and skills and before he and the organization can thoroughly assess the fit. To avoid such a scenario, LiveOps followed our suggestion to place new workers in the top 10% for their first 300 calls, regardless of performance. This lets them make use of call-by-call feedback and gain an understanding of the ranking system before deciding whether to stick around. Our preliminary analysis suggests that the strategy is effective: New workers are more motivated, and good performers among them stay with the firm longer than in the past.

The system must measure quality as well as quantity. Restaurants can do this by tracking tips. A waitress who has too much work and does not pay enough attention to individual customers will see her ranking slide because of low tips. She’ll then be assigned fewer tables and can spend more time with each one, giving her a chance to bring her ranking back up. This prevents a race to the bottom and allows everyone to find and maintain an optimal performance level in the long run.

We believe that the opportunities for implementing winners-take-all systems are plentiful. It’s easy to imagine the model’s working in virtually any retail or transaction-oriented service business. Wherever there is a best location or a preferred shift, and as long as performance can be evaluated and ranked, companies can realize productivity and profit gains by shifting work to their best performers. And managers, freed from many of the tasks involved in appraising employees, can focus more attention on activities that really count—marketing, promotion, and other efforts that add to a business’s long-term value.