Straight answer: you research salary ranges by pulling numbers from three or four sources that don’t talk to each other (Glassdoor, LinkedIn Salary, a government wage survey, and one human being who works in the industry), then you throw out the outliers and anchor on the top third of what’s left. Most people stop at one website and wonder why their number gets laughed at. The websites are a starting point, not the answer.
Why one website will always let you down
I’ve sat across the table (well, across a Zoom screen, these days) from clients who walked into salary talks armed with a single Glassdoor screenshot and nothing else. It’s the single biggest mistake I see, and I made it myself years ago before I knew better.
Here’s the problem with relying on one source. Glassdoor data is self-reported, often outdated by a year or two, and skewed toward people who felt strongly enough (usually negatively) to bother submitting a number. LinkedIn’s salary tool is better for volume but leans heavily on people who’ve listed their current pay in their profile, which in the UK is still a small minority. Payscale asks you to fill in a survey before it shows you anything, which means its sample is self-selecting in a different way again.
None of these are wrong. They’re just partial. Used alone, any one of them gives you false confidence, and false confidence in a negotiation is worse than no confidence at all, because you’ll state a number with certainty and then crumble the moment someone pushes back on it.
The four sources I pull from
When I’m helping someone prepare for a negotiation, whether that’s a salaried role or a freelance day rate, I ask them to gather data from four different types of source before we talk numbers.
- Aggregator sites: Glassdoor, LinkedIn Salary Insights, Indeed Salaries, Payscale. Use at least two, not one.
- Official wage data: in the UK, the Office for National Statistics publishes the Annual Survey of Hours and Earnings (ASHE), which is free, based on tax records rather than self-reported guesses, and broken down by occupation and region. In the US, the Bureau of Labor Statistics Occupational Employment and Wage Statistics does the same job.
- Recruiter salary guides: Hays, Robert Half, Michael Page, Reed, and Totaljobs all publish annual UK salary guides, usually free to download, built from thousands of live placements rather than surveys. These are gold because they reflect what companies are paying right now, not what someone claims they earn.
- A real conversation: one person doing the job, or close to it, who will tell you honestly what they’re on or what they’ve seen offered. This is the source that catches everything the spreadsheets miss.
I’ll be honest about the bit nobody likes to say out loud: the recruiter guides and the human conversation matter more than the flashy salary calculator websites. The calculators are what everyone uses because they’re free and instant. The guides and the conversation are what change the outcome, and almost nobody bothers with them because they take more than five minutes.
A real example: Sarah’s £6,000 gap
A client of mine, I’ll call her Sarah because that’s not her name, was offered a marketing manager role at a fintech firm in Manchester. The offer was £48,000. Her Glassdoor search for “marketing manager Manchester” showed a range of £38,000 to £62,000, which told her precisely nothing useful. It’s too wide to anchor on.
We narrowed it down. LinkedIn’s salary tool, filtered to fintech and companies of a similar size (80 to 150 employees), showed a tighter band of £47,000 to £58,000. The Hays UK Salary Guide for that year listed marketing manager roles in financial services at £45,000 to £60,000 outside London. Then Sarah messaged a former colleague who’d moved into a near identical role at a competitor eighteen months earlier. That colleague, over a fifteen minute call, told her the going rate for that seniority in Manchester fintech was “comfortably mid fifties, and don’t take less than 52.”
Four sources. Three of them roughly agreed. The fourth, the human one, gave her the confidence to ignore the wide Glassdoor range entirely and open at £57,000. She settled at £54,000, six thousand pounds above the original offer, plus an extra day of annual leave she hadn’t even asked for going in. That’s not a fluke. That’s what happens when you stop treating one website as gospel.
The uncomfortable truth about published salary ranges
Since 2023, and expanding further through 2026, a growing number of US states (Colorado, California, Washington, New York, Illinois) and several UK employers voluntarily now publish salary bands on job adverts. On paper this sounds like a gift for anyone researching pay. In practice, the range is often deliberately wide enough to be almost meaningless. A posting that says £45,000 to £70,000 for the same title tells you the company wants to stay on the right side of a transparency law, not what they intend to pay a specific candidate.
Companies build these ranges to cover every possible hire, from a junior stretching into the role to a director-level candidate they’d love to land if the budget stretched. The number they offer you sits inside a much narrower band based on your specific experience, your internal comparators (what the person next to you is already earning), and how badly they need the role filled by a certain date. Nobody puts that narrower band in the job advert, because it would expose exactly how little wiggle room most roles have.
So when you see a published range, treat the bottom third as the floor for someone with minimal relevant experience, the top third as reserved for people they’re trying to poach from a competitor, and aim your research at figuring out where you sit in the middle third. That’s the number worth building your case around.
Step by step: the research process I’d use this week
If I were prepping for a negotiation tomorrow, here’s exactly what I’d do, in order, over roughly three to four hours spread across a couple of evenings.
- Step 1: Search the job title on Glassdoor, LinkedIn Salary, and Indeed. Write down the range from each, plus the sample size if it’s shown (a range built on 400 data points is worth more than one built on 6).
- Step 2: Check the ONS ASHE dataset (UK) or BLS OES data (US) for the specific occupation code closest to your role. This gives you a median that isn’t self-reported, which is your reality check against the aggregator sites.
- Step 3: Download the newest annual salary guide from at least two recruitment firms that place in your sector. Hays, Robert Half, Michael Page, Reed. These are free and updated yearly.
- Step 4: Look at three to five live job adverts for near identical roles at similar sized companies in your city or region, especially ones that publish a band. Note where in the band roles requiring your level of experience seem to sit.
- Step 5: Message two people. One who does the job now, one who’s interviewed for it recently elsewhere. Ask them directly what they’d expect to be offered, not what they earn if that feels too personal. People are far more willing to share a hypothetical number than their own payslip.
- Step 6: Take every number you’ve gathered, throw out anything more than 15% above or below the cluster of the others, and build your ask from the upper half of what remains.
That’s it. It’s not complicated. It just takes longer than typing a job title into one search box, which is precisely why most people skip half of it.
Company specific research, not just market research
Market rate tells you what the job is worth broadly. It doesn’t tell you what this specific employer will pay. For that, check Glassdoor’s company specific reviews and salary submissions (filter to just that employer, not the general job title search), look at Companies House filings if it’s a UK limited company (their most recent accounts show wage bill trends and whether headcount costs have been rising or squeezed), and check whether the company has had recent layoffs or a funding round, both of which change what they can stretch to right now regardless of what the market average says.
I once turned down a speaking engagement based purely on “market rate for a keynote speaker” only to find out afterwards, through a mutual contact, that the same event had a budget line nearly double what I’d assumed and had already paid another speaker at that level the previous year. Market research told me the ceiling for the industry. It told me nothing about that particular client’s actual budget. Those are two different numbers and conflating them costs people money constantly.
For freelancers and contractors, the research looks slightly different
If you’re pricing a day rate or project fee rather than a salary, the aggregator sites are far less useful because self-employed pay data is thin and inconsistent. I wrote about how I fixed my own pricing anxiety in how I used AI to finally set rates I don’t feel sick about charging, and the short version is that I stopped guessing and started building a spreadsheet of every rate I’d been offered, quoted, or heard about from other consultants at my level, then priced against that instead of a vague sense of what felt fair. Freelance rate research is really just salary research with worse public data, so you have to lean even harder on step 5, the direct conversation, because there’s no ASHE equivalent for consultants.
If you’re earlier in a freelance career and still working out what the market will bear at all, it’s worth reading the one freelance tip that changed my income, because pricing research only works once you’ve fixed the underlying confidence problem that makes people undercharge regardless of what the data says.
What to do with contradictory numbers
Your research won’t always line up neatly. Sometimes ASHE says one thing, LinkedIn says another, and your contact in the industry says a third figure that doesn’t match either. When that happens, weight the numbers by how close they are to your actual situation. A national government average is broad and slightly stale but trustworthy in direction. A recruiter guide reflects live placements but skews toward roles that recruiter agencies fill, missing anything hired purely through internal networks. A single friend’s number is specific but could be an outlier in either direction.
My rule of thumb: if three out of four sources cluster within about 10% of each other, trust that cluster and treat the fourth as the outlier. If all four disagree wildly, that’s usually a sign the job title itself is too broad or too varied by seniority for a single number to mean anything, and you need to narrow your search terms further (add seniority level, add sector, add company size) before you try again.
Timing your research matters more than people think
Salary guides get published annually, usually in January. Government wage data often lags by six to twelve months by the time it’s released. If you’re negotiating in a fast moving sector, tech especially, a salary figure from eighteen months ago can already be out of date, particularly after a period of sharp hiring or sharp cuts. Always check the publication date on whatever source you’re using and mentally discount older data upward or downward depending on whether the sector has been hiring aggressively or freezing.
This is also where job search support can earn its keep rather than just adding noise. If you’re deep in an active search and don’t have hours spare to cross reference four data sources every time a new role comes up, a job search virtual assistant can be told exactly what to pull for each role you’re considering, so the research happens in the background while you focus on interviews.
Set your number, then prepare to defend it, not just state it
Research gives you a number. It doesn’t give you the nerve to say it out loud without flinching. Write down, in one sentence, why your number is fair, based on the specific sources you used, not a vague sense of deserving more. “Based on the Hays 2026 salary guide and three comparable roles currently advertised in this city, £56,000 is the market rate for someone with my experience” lands very differently to a shrug and “I was thinking maybe 56.”
What most people get wrong going into salary talks isn’t the research at all, it’s the expectations they’ve built around what negotiating even looks like, which is worth reading up on separately in expectations versus reality for job seekers before you get anywhere near a number.
Frequently asked questions
How many sources should I check before I trust a salary figure?
At least three, and ideally a mix of types: one aggregator site (Glassdoor or LinkedIn), one official wage dataset (ONS ASHE in the UK, BLS OES in the US), and one direct conversation with someone in a similar role. Two sources of the same type agreeing with each other still isn’t enough, because they often share the same flawed self-reporting problem.
Is it worth paying for a salary report instead of using free tools?
Rarely. The recruiter salary guides from Hays, Robert Half, Michael Page, and Reed are free, updated yearly, and built from actual placements rather than surveys, which usually makes them more reliable than a paid report from a smaller research firm. Save your money and spend the time instead.
Why do published salary ranges on job adverts seem so wide?
Because companies build them to cover every candidate they might realistically hire, from junior to near director level, and to comply with pay transparency laws without revealing their actual budget for a specific candidate. Treat the middle third of a published range as your realistic target, not the top.
Should I tell people what I earn to get better data from them?
You don’t need to. Most people will happily tell you what they’d expect a role to pay hypothetically, even if they won’t share their own payslip. Ask “what would you expect this to pay” rather than “what do you earn” and you’ll get far more honest answers with far less awkwardness.
Official documentation
Related reading: Virtual Assistant Employment: The Real Salary, Why Half Quit in Year One, and What Works and How Do You Negotiate a Higher Salary During a Job Offer?.
