For Darwin Hanson, it was an awfully close call. The independent analytics consultant was helping a client—a big-name aerospace company—crunch salary data when suddenly he noticed that the numbers just weren’t adding up.
“What the data showed was that a person who was just hired as an engineer was eligible for a larger salary than a person who had been fully competent in the position for two to three years. Up until that point, nobody had caught the error,” recalls Hanson. The mistake—although detected just in the nick of time—still cost the company nearly $20,000 in lost productivity and consultation fees.
Welcome to the fast-paced world of compensation analytics. These days, countless compensation survey vendors, from Mercer to Aon Hewitt, are under unprecedented pressure to produce reams of data for analytics purposes. Companies then purchase these data packages to determine factors such as the top pay level for a seasoned engineer, bonus payouts for new sales reps or compensation packages for senior-level managers.
The problem, however, is that a single miscalculation or error in data can cost a company hundreds of thousands of dollars.
The problem of data quality is not confined to HR executives using analytics and benchmarks to manage their workforces. A 2009 survey by Gartner Research found that poor data quality costs companies an average of $8 million per year. Data governance experts bemoan the lack of data consistency for the waste it can create in business processes such as customer service. A flawed smartphone mapping application led Apple to make changes in its executive ranks after iPhone users complained about errors and navigational mishaps.
In the HR function, faulty data costs corporations in more ways than consulting fees. High attrition rates, poor employee morale and legal liabilities are the potential by-products of data typos, oversights and inaccuracies.
Part of the problem is the process survey vendors rely on to crunch data. Typically, a vendor will send out a request for information—a call for survey submission—that asks a client to fill out an extensive number of documents detailing variables such as employee salary, bonuses, compensation and benefits. These datasets are then combined with information provided by competing companies to produce important industry benchmarks and market prices. In the end, a survey vendor may be responsible for analyzing upwards up 60,000 records at a time, says Hanson. All of which leaves plenty of room for mistakes to be made by participating companies and vendors alike.
The Fallout of Faulty Data
Another data landmine that can lead to financial loss: varying job definitions. For example, a survey vendor may request that a company submit salary information on its accountants. But one company’s accountant may sometimes be another’s glorified bookkeeper. Failure to flag these semantic nuances can result in misleading information.
So too can parsing data to its greatest granularity backfire. Many companies make the mistake of demanding that survey vendors crunch numbers to the nth degree such as the market price for a software sales rep, under the age of 50, with 10 years’ experience and a degree from MIT. “The more you slice and dice your data, the smaller the sample size,” warns Lindsay Scott, founder of Lindsay Scott & Associates, a compensation consulting firm. “And in statistics, if the sample size is very small, the information that data suggests may not be correct.”
Scott would know. He recently unveiled the Market Data Quality Assurance Tool – a computer model that provides independent analysis of compensation vendor survey data. The software works by using dozens of checkpoints to examine the accuracy and quality of the data being analyzed.
But the cost of incorrectness extends far beyond paying your bookkeeper an accountant’s wage. Workforce morale, industry compliance, customer attrition—they all stand to suffer from managers making key decisions based on faulty data.
For Hanson, some $20,000 in lost productivity and consultation fees was a fraction of what his client would have had to pay if the error in salary data hadn’t been detected in time.
“If it didn’t get caught and HR posted these numbers, then suddenly all of the employees would have become suspicious,” says Hanson. “Then it’s a worst case scenario all around because when your employees don’t trust you, they’ll start looking elsewhere for employment no matter what you do.”
Scott also points out that the flip side to paying employees the wrong salary is just as costly. “When the data’s not correct and you’re not paying people the way they should be paid, one of two things happens,” he says. “Either you’re paying them more than you should and they’ll never leave, or you’re paying them less than you should and they’re always leaving. Neither one of these things you want.”
In fact, The Saratoga Institute, an HR consulting organization, suggests that the average internal cost of turnover for an employee is a minimum of one year’s pay and benefits or a maximum of two years’ salary.
Incorrect data can also prompt HR managers to make the wrong hiring decisions. For example, an HR manager relying on poorly collected and analyzed compensation data may not offer candidates a package that’s commensurate with their market value.
“If you have data that is for one reason or another either incorrect or misleading, and you use it for a hiring decision and don’t get that candidate because the data is misleading, then that’s a problem,” says Scott.
In the end, though, perhaps the greatest price companies pay for faulty data is in lost productivity. To avoid errors, more and more companies are working hand in hand with survey vendors to ensure they’re providing the right information, that job descriptions match up and that data relating to pay scales and compensation packages actually make sense. It’s necessary due diligence but a process that demands months’ of dedicated work. As for those errors that fall through the cracks, companies can expect to pay for everything from disgruntled workers to the helping hand of a statistician or two.
Cindy Waxer is a Toronto-based freelance journalist and a contributor to publications including The Economist and MIT Technology Review. She can be reached at firstname.lastname@example.org or via Twitter @Cwaxer.