Salary benchmark vs salary survey: why the numbers don't match what you earn

A friend reads the latest Robert Half salary guide and sees "Senior Software Engineer, Sydney: $180K–$220K." She earns $155K. She asks her manager for a $35K raise, citing the guide. The manager pushes back: "Our band tops out at $185K, and you're already at the top." Both numbers are real. Both come from credible sources. Why don't they agree?

The short answer is that a salary survey and a salary benchmark are different tools, built from different data, answering different questions. Treating one as the other is the most common mistake people make in salary conversations.

This guide explains what each actually is, the five structural reasons they diverge, and how to read each correctly when you're negotiating.

What a salary survey is

A salary survey is a periodic publication — usually annual — produced by a consulting firm, recruitment agency, or industry body. The Robert Half guide, the Hays salary guide, and the Michael Page salary report are the big three in Australia.

The methodology is roughly this:

• Survey thousands of professionals and employers in a given period (usually the months before publication)
• Aggregate the responses by role and location
• Publish ranges with low / median / high figures for each role

Surveys are useful for one thing in particular: understanding the broad shape of a market. They tell you whether "senior data engineer in Brisbane" is a $130K role or a $200K role at the level of magnitude. Past that, they fall apart fast — for reasons we'll get to.

What a salary benchmark is

A salary benchmark is a point-in-time market estimate for a specific role. The inputs are different:

• Live job listings (advertised salary ranges)
• Industry-body wage data (ABS, FWO)
• Internal compensation data (where available)
• Recent offer data from talent sources

A benchmark answers a narrower question: "What is the market paying right now for this specific role in this specific location, for someone at this specific level of seniority?" When done well, it includes source citation so you can see which listings or surveys it pulled from.

The distinction matters because a salary benchmark is intended to be acted on. A survey is intended to be read.

The five reasons the numbers diverge

1. Surveys lag the market by 6–18 months

Most salary guides are published annually. The data collection for the January 2026 guide probably ran in September–November 2025. The respondents reported on their then-current salary, set in previous review cycles — which could be from January 2025 or even earlier.

In a stable market, that lag doesn't matter. In a moving market — like Australian tech in 2022–2024 — the lag is the difference between accurate and 20% wrong. By the time you're reading the guide, the market has already moved past the numbers in it.

A live benchmark uses listings posted this week. The lag is days, not quarters.

2. Survey respondents are self-selected

Who fills out a 25-question salary survey? People who care about salary data. People who think they're underpaid (looking for ammunition). People who recently changed jobs (proud of the result). People who follow the agency on LinkedIn.

People who aren't in those buckets — the median worker, the long-tenured employee, the contractor — don't respond. The distribution of respondents isn't the distribution of all workers. Surveys correct for this with weighting, but the correction is imperfect.

A benchmark using listing data has a different bias (it only sees roles employers are currently hiring for), but it's a different bias — meaning when both agree on a number, you have stronger evidence than either alone.

3. Role titles mean different things at different employers

"Senior Software Engineer" at Atlassian, at a 50-person startup, and at a big-four bank are three different jobs. The Atlassian role assumes a level of system-design ownership; the startup role might be 70% IC, 30% manager; the bank role might be a glorified mid-level title with extensive process overhead.

Surveys aggregate all of them under "Senior Software Engineer." The headline number is an average of three quite different markets.

A good benchmark accounts for this by reading the actual job description — required skills, scope, leadership expectations, tech stack — and matching against similar-shape roles, not just same-title roles.

4. "Sydney" hides huge variance

A Sydney listing could be a CBD-based fintech, a Western Sydney council, or a fully-remote role tied to a Sydney HQ for tax purposes. These three pay very differently — sometimes $50K+ apart.

Most surveys collapse to capital-city granularity at best. A live benchmark can see the actual employer, the actual remote/hybrid policy, and the actual industry classification on each listing.

5. Surveys don't reflect your salary band

This is the most important point. A survey reports market data. Your salary band is your employer's policy. They're related but not the same.

Even when the market broadly supports $180K, your specific employer might have a band that tops out at $165K because:

• They benchmark internally against companies in their tier (e.g., "we pay 50th percentile of ASX-listed peers")
• They lag the market by design (correcting for market movement during cycle)
• They have headcount caps and budget constraints
• They're optimising for retention over hiring (different lever profile)

A survey can tell you "the market is at $180K." It cannot tell you "your employer's band is at $165K and they won't break it." That conversation requires understanding where in the band you sit and what percentile of market your employer targets (see salary percentiles explained).

So which should you use?

Use both, but for different purposes:

• Salary survey — for orientation. "Is this role in the ballpark of $150K or $250K?" Surveys are great for ruling out obviously wrong numbers.
• Salary benchmark — for action. "Should I ask for $175K or $195K for this specific role at this employer in this market right now?" Benchmarks are what you bring to a real conversation.

The mistake is bringing a survey number into a salary conversation and treating it as a benchmark. Your manager will see straight through it — they have access to the same surveys and know all the reasons not to trust them.

The opposite mistake is bringing a benchmark to a question survey-shaped questions (e.g., "Is this industry well-paid in general?"). A benchmark is too narrow to answer broad questions.

What to do with this

Two practical takeaways:

1. When researching salary, read at least two sources. If Robert Half and a live benchmark agree within 5%, you can act on either with confidence. If they disagree by 20%, dig in — usually one of the five reasons above is the culprit.

2. Before any salary conversation, get a benchmark — not just a survey. Cross-check free salary sources — recruiter guides, crowd-sourced sites, official data — so you triangulate rather than anchor on one. EvenBetter Salary does this triangulation automatically; see our methodology for the source list and how we weight each one. Then run a live benchmark of your role so you walk in with the right number.

The numbers in the salary guide aren't wrong. They're just aggregated to the wrong level of resolution for the question most people are asking.