Paying with confidence: How technology is transforming compensation practices
Machine learning has turbo-charged traditional compensation surveys to give employers better insight into competitive pay practices. Alys Reynders Scott, Salary.com’s CMO, provided an update on the world of salary surveys as it pertains to her own company, and beyond, in a recent conversation with Employee Benefit News. Edited highlights follow.
Employee Benefit News: How do you use Salary.com’s data resources for your own internal purposes?
Alys Reynders Scott: We use our tools and expertise to set salary ranges for all levels across our organization. We fall pretty plainly within the ranges our data tells us we should be, and we communicate openly with our employees about our salary ranges for any roles. We also communicate our pay philosophy. We’re pretty mindful of that because we are a compensation company, but also because it’s an important way to retain employees.
EBN: And to attract employees…
Scott: Right, we also use our data extensively in our recruiting process. We’re in Waltham near Boston, and we have around 100 people here plus 70 in other places around the world. Boston is a really tight labor market, especially for technology jobs. So for us in a 3.9% unemployment rate geography and in the tech sector, competitive pay and getting the right offer out quickly are essential for filing an open position.
EBN: In that environment, is it tempting just to err on the side of being a little high, and not analyzing the numbers to death?
Scott: No, employers need speed and accuracy, to look at compensation in a more granular way. We look at competencies for a role and we have a database of thousands and thousands of job descriptions. So we can rely on our own job descriptions, we can rely on the competencies associated with those and how they attribute pay to those competencies. But we also look at some compensatory factors like graduate degrees, certifications and other elements as they relate to pay.
EBN: What else goes into the mix when you’re trying to come up with the right package?
Scott: We also look at factors like long- and short-term incentives — what kind of roles in an organization are eligible for them. And you have to consider how different demographics respond to them. For example, most of my team are millennials or younger. For this population, having a quarterly bonus model can be really important, instead of just at the end of the year.
EBN: How do you factor in intangible compensation elements that might make one employer more attractive than another, even if the pay is essentially the same?
Scott: Yes, that’s important too. If you’re a hip start-up software company, your whole office is all glass, there’s free food in the kitchen and 7,000 coffee machines, you’ve got a collaborative and innovative culture, that helps but it only goes so far. Today that’s what many candidates already expect.
EBN: What else can your system tell employers besides market compensation numbers?
Scott: We have metrics that can tell you which jobs are “hot,” based on how rapidly pay for those positions is going up. It might be a job that didn’t even exist 10 years ago, like a digital marketing strategist. If it’s a hot job, you might decide you need to go higher than the 50th percentile of compensation to be competitive.
EBN: How can employers use geographic splits of the data to know what’s competitive?
Scott: You can look not only at data for your own location, but for others. You also can start with a number for a specific job where you are, then re-price it for another market. A widget in our analytics lets you look at a job, for example, in San Francisco, and price it for Boston, based on differences in the cost of living. If you’re planning to open a new office, you can do some modeling.
EBN: But that doesn’t mean you can just open an office in a less expensive market and assume that you’ll find who you’re looking for there at a lower price, can you?
Scott: That’s right, you have to titrate a decision between the cost of the role and the availability of the talent. So you want to pick a market that’s going to be cheap enough, but you want to make sure that you have a large enough candidate pool. Plus you have to look at the implications of the cost of housing, not just the prices in isolation. There are a lot of jobs in San Francisco and Boston but the cost of housing is so high that people have to live far away. So as an employer you have to think about whether paying some commuting expense needs to be part of the compensation package.
EBN: How do “pay ban” laws in states like Massachusetts, where employers can’t ask job candidates for their salary history, affect use of compensation survey data?
Scott: Pay bans are moving across the country incredibly quickly. You can still ask people about their pay expectations, and have a conversation about your pay policies. When you make an offer, the number shouldn’t be a huge surprise to the candidate.
EBN: But that doesn’t prevent a candidate from making a counteroffer…
Scott: That’s right, and negotiation is often necessary. But you’re in a stronger position when you have good data to support your offer and can explain how you got to that number.
EBN: So many jobs now aren’t so neatly categorized? How can survey data be made useful for jobs that aren’t common?
Scott: We have cultivated our data from more than 300 surveys, and we analyze it with machine learning and artificial intelligence. The value is your ability to make sense of the data, apply the data. Our analytics lets you look at different job families, price jobs across families, across different focus areas, across different functions in an organization. The flexibility of our analytics that takes you out of really a very rigid data architecture. This is important when we live in an environment where there are blended jobs, to be able to look at multiple jobs requiring different areas of expertise, and determine what’s appropriate.
EBN: What happens if an employer is in a small market where there aren’t a lot of people in doing a particular job, so that local survey data would have a pretty big margin of error?
Scott: This is an area where machine learning comes into play. We have a series of algorithms that look at all of our data over a 30-day period and continuously refresh the data. So what happens in a city where we have only, say, 12 specific data points for an individual job? We can take data from other locations, and transform it with cost of living and adjacent skill sets and competencies for the role taken from other jobs, and apply that logic to create a larger pricing sample to maximize the accuracy.
EBN: Once somebody is on board, you’ve complied with local laws that are trying to end disparate pay patterns based on past discrimination, what happens if the market tightens and you just have to pay new hires more than people you hired a few years ago in a softer economy?
Scott: That’s where predictive analytics come in, to look at things like flight risk — proactively identifying cases where someone might be underpaid relative to the market. There is an equity issue now with pay compression at many companies — some people who were hired during and right after the financial crisis aren’t earning much more than some people hired more recently for similar jobs. Some employers will look at the data and grapple with the issue, even if they don’t believe the flight risk is high.