Pay Transparency and the Problem of Old Salary Data

When a law makes you publish a number, you find out quickly how much you trusted it.

That's roughly what happened with pay transparency. As more places required salary ranges on job postings, most talent teams did the sensible thing: they took their compensation benchmarks, turned them into ranges, and put the ranges on the posting. What few of them asked was how old the data behind those ranges was by the time a candidate read it.

Trace it through and the answer is older than most people would guess. Compensation surveys run on a fixed yearly cycle. Participating companies submit data over a window that usually closes in the spring. The vendor, whether that's Radford, Mercer, or Willis Towers Watson, cleans it up, adjusts for geography and role scope, and publishes in the fall. Then your comp team reviews the results, runs an equity analysis, builds bands for the next cycle, gets them approved, and loads them into the ATS. Suppose you post a job the following March. The pay data behind that range was collected fifteen to eighteen months earlier.

In a steady market that doesn't matter much. In a market where pay for some roles has jumped in a couple of years, or where the job family barely existed when the survey was taken, it means you're mispricing the role for a whole hiring season. None of this is the vendor's fault. A survey records what companies paid. It can't tell you what companies are offering right now.

The candidate has newer data than you

Meanwhile the candidate is looking at something much more recent. Blind, Levels.fyi, LinkedIn Salary, and Glassdoor all collect pay data close to real time, and a senior engineer considering your role can look up the last few months of reported offers in their market before deciding whether to apply. In specialized technical fields, people share offer details with each other quite precisely: base, equity, vesting, sign-on. So your hiring manager is working from a benchmark that's a year or more old, and the candidate across the table is working from last week's numbers.

Consider what that does. Some good candidates see your range, decide it's under market, and never apply. Others apply anyway, spend three weeks in your process, and find out at the offer stage that the numbers don't work. By then you've spent a lot of hiring manager time on a process that was always going to fail at the last step.

Neither shows up clearly in the usual metrics. The people who opted out just look like low application volume. The late misses show up as offer declines that look like candidate experience failures, when the real cause was a price set before the search began. And since candidate drop-off at the offer stage has no standard line item in the recruiting dashboard, hardly anyone goes looking.

Asking the candidates

If the market has moved, the people who know by how much are the candidates. They know what they earn, what it would take to move, and what else they're being offered. The trouble is that this usually comes out late, in a negotiation, one candidate at a time, and it isn't written down anywhere you could add it up.

That's one reason compensation is a standard part of the screening call in Asendia AI. We make software that phones applicants soon after they apply and interviews them. Among other things it asks what the person earns now, what would make a move worth it, and what other conversations they're in. Because the questions come in the same order every time, and the follow-ups push for real numbers instead of vague talk about market rate, the answers can be compared. They go into your ATS with the rest of the screening notes.

Suppose you ran fifty of those conversations for one role family over a month. You'd have a fairly current picture of that market: whether your range is putting people off, whether candidates keep weighing you against one particular competitor and by how much, and which parts of the package they care about most, base or equity, cash or flexibility. An annual benchmark can't tell you that, because it describes last year.

Pay transparency made you put a number on the page. It didn't make the number current. To see whether yours is stale, pull the last two quarters of declined offers, look at the ones where the stated reason was another offer, and estimate the gap between what you offered and what the candidate probably got. If the gaps cluster in one role family or level, you're looking at benchmark lag showing up as lost hires, and the information that would have caught it was already passing through your screening calls.

Ready to transform your hiring strategy? Schedule a Demo with our founders today!: https://asendia.ai/talk-to-founders