This article originally appeared on HBR.org.

How much is your top salesperson worth? Your star engineer? Your best marketer? Everyone knows that some people get better results than others. But the best aren’t just a little better than the rest — they’re typically a lot better. Bain & Company’s research, discussed here, suggests that top performers are roughly four times as productive as average performers. Sometimes the difference is far greater. For example, the best sales associate at Nordstrom sells at least eight times as much as the average sales associate at other department stores.

Given the sizable differences between the best and the rest, a company with a higher percentage of top performers will naturally tend to outdo its rivals. The reason is that it has higher human capital productivity (HCP), which we know correlates closely with financial results. Raw talent isn’t the only determinant of HCP, but if you don’t have the “A” performers you need, none of the other factors will make much difference. So improving your overall talent level is the first step toward higher HCP.

How can a company raise its skill level—and in particular, how can it increase its proportion of top talent? Our research and experience at Bain suggests three keys.

Assess your talent pool. You can’t know the magnitude of the task you’re facing until you know exactly where your human-capital strengths and weaknesses lie. AllianceBernstein (AB), a $3 billion asset management company based in Manhattan, rates each of its 3,700 employees every year on both performance and potential. The senior team at AB spends several days together each year cross-calibrating the two sets of ratings across the entire company, so it always knows where its top performers are. One caution: performance and potential both matter, but performance should count for a lot more. Performance is real; it can be measured objectively. Potential is always subjective, and may never be realized. So pay close attention to your high performers. If they’re a tiny fraction of your overall talent pool, you know you have a problem.

Control your pipeline. Whenever possible, avoid relying on executive search firms as the primary source of new talent. A company looking for more A players needs to know first-hand about the talent that is available, and it needs to do its own recruiting. One well-known Silicon Valley firm relied heavily on headhunters to find engineers for many years. The result? It got engineering candidates who couldn’t land jobs at Apple, Google, Facebook, and other A-list companies in the valley. Only the arrival of a new CEO led to a change in the policy — and an eventual uptick in the company’s talent base and performance.

Have only high-performers conduct the interviews. It’s a sad truth about human beings: B and C players can’t always identify top performers, and may feel threatened by them if they do. Average performers look for congeniality, the ability of an interviewee to fit in. They don’t always look for — and they may not favor — someone who seems likely to raise the bar or make them look bad. One other tip: involve as many line managers in the interviews you can. Many jobs these days are highly technical, from software development to running a copper mine. Interviewers from HR aren’t usually capable of judging someone’s technical skills.

A company, like a sports team or a symphony orchestra, has to do a lot of things right to reach the pinnacle of performance. But if it doesn’t have a healthy share of those exceptional individuals who do so much better than everybody else, then it’s out before the race even begins.

Michael C. Mankins is a partner at Bain & Company. He is based in San Francisco and heads Bain's Organization Practice in the Americas.