Most people walk into a salary conversation with a number they half believe. They looked at one website, remembered what a colleague said two years ago, added a bit for confidence, and hoped nobody asked how they got there. Then a recruiter asks for their expectations in the first ten minutes and the whole negotiation is decided by a guess.

Research fixes this, and it takes about ninety minutes. The aim is not a single magic figure but a range you can defend with sources, adjusted for your market and your seniority, with a floor you have decided in advance. This guide covers why public salary data is so noisy, which sources are worth your time, how to turn conflicting numbers into one range, what remote work does to it, and how to state the range without cornering yourself.

Why is salary research so unreliable?

Because the same job title covers wildly different jobs. A marketing manager can run a team of nine or be the entire marketing department, and both appear in the same average. Titles inflate in small companies and deflate in large ones, which means a title-based average blends people whose actual responsibilities barely overlap.

Then there is how the data is gathered. Self-reported sites are skewed by who bothers to submit, and totals often mix base pay with bonus and equity without saying so. Government statistics are reliable but lag by a year or more and use broad occupational groups. Job advert ranges are current but frequently wide, sometimes deliberately, and the top of an advertised band is usually reserved for someone with more experience than the person reading it. None of that makes research pointless. It means you should never rely on one source, and you should always ask what a number includes, since wage data can mean base pay, total cash, or total package depending on who compiled it.

The sources worth using

Use several, then look for the overlap rather than the extremes.

  • Job adverts with published ranges. The most current signal available, and legally required in a growing number of places. Collect ten for your role and level in your area.
  • Official occupational data. Government sources such as the national statistics office publish median wages by occupation and region, and in the United States the Bureau of Labor Statistics does this annually. The occupation profiles at O*NET Online pair wage bands with the actual tasks of the job, which helps you check you are comparing the right role.
  • Aggregator sites. Tools such as PayScale let you filter by experience and location, which matters more than the headline average.
  • Industry-specific data. In technology, Levels.fyi breaks compensation into base, bonus, and equity by level, which is the correct way to read any package with stock in it.
  • Recruiter salary guides. Published annually by staffing firms, useful for direction and slightly optimistic by design.

Ignore any single number that sits far outside the rest. One outlier is usually a mislabelled job, a different city, or somebody counting equity as salary.

How to turn data into your own range

Work in four steps. First, gather your ten advert ranges and note the midpoint of each, then find the middle of those midpoints. That is your market centre for the title, before any adjustment for you.

Second, place yourself. Compare your years of relevant experience, the size of what you have owned, and any scarce skill against the level described in those adverts. Being a strong fit puts you above the centre; stretching upward into a bigger role puts you at or slightly below it. Be honest here, because the point of the exercise is a number you can defend when someone asks what you have done.

Third, build the range. Set your target at your adjusted market figure, your ask about 10 to 15 percent above it, since almost every offer is negotiated downward from where it starts, and your floor at the number below which you would decline. Write the floor down before any conversation happens, because floors set under pressure are always lower.

Fourth, convert to total value. A base salary is not comparable across companies until you add the retirement match, bonus structure, health coverage, and paid leave, so keep the whole package in view using the checklist in how to evaluate a job offer. Two offers 8,000 apart in base pay are often equal once benefits are counted, and occasionally the lower one wins.

How do location and remote work change the number?

Location still moves pay substantially, because employers price against the local market for the role. The same job can differ by 20 to 40 percent between a capital city and a smaller regional market, so filter every source by geography before you compare anything. An unfiltered national average is the least useful figure you can carry into a negotiation.

Remote roles complicate this in a specific way: you need to know whether the company pays for the location of the work or the location of the worker. Some pay one global band for a role, which is excellent news if you live somewhere cheaper. Others apply location bands and will adjust an offer, or your future pay, to where you live. Ask which policy applies before you name a number, and ask what happens if you move. For fully remote roles the relevant market is often the company's headquarters, which is why remote work can be a large effective raise, and it is worth checking the market rate for both locations while you are searching for legitimate remote jobs.

Asking real people (the highest quality data)

The best data is not on a website. People who do your job in your city know what the market pays, and many will tell you if you ask in a way that is easy to answer. Instead of asking what someone earns, ask what range they would expect a role like this to pay, or what a fair band is for this level. That is a question about the market rather than about their payslip, and it gets answered far more often.

Recruiters are the most efficient source of all, because they see live offers weekly. A specialist recruiter in your field will usually tell you the band for your level in exchange for nothing more than a conversation, and it is worth having that call even when you are not actively applying. Former colleagues who moved recently are similarly valuable, since they have seen a real number in the last twelve months rather than an average from last year. This is one of the practical reasons to keep contacts warm, as covered in networking for a job. Talking about pay openly is also becoming normal rather than impolite, and the people who benefit from silence are rarely the employees.

How do you use your range in an application?

When an application form demands a figure, give a narrow range rather than a single number, and put your target near the bottom of it. If a recruiter asks early in a call, it is fine to say you would rather understand the scope first and ask what band they have for the role, since they almost always have one. If they insist, give your researched range and add that you are flexible depending on the full package. Never invent your current salary, because it is easy to check and the cost of being caught is the offer.

Keep the research and the negotiation separate in your own head. Research tells you what the number should be; the conversation is where you defend it, and the wording matters, which is why the scripts in how to negotiate salary are worth rehearsing out loud. The same range also becomes your reference point once you are inside the company and asking for more, so keep the file and update it annually before your review, using the approach in how to ask for a raise. Ninety minutes of research is usually the best paid hour and a half in the whole job search. More guides are in the article library.