Why look up ranges before you talk to anyone
Walking into a process with no sense of the pay range puts you at a disadvantage that has nothing to do with skill. You cannot tell whether a recruiter's number is generous or low, you cannot judge whether a role is worth the time of a long interview loop, and you cannot give a sensible answer when someone asks what you are looking for. A bit of homework fixes all three problems.
This guide is about research, not tactics. It does not tell you what to ask for or how to push a number up. The goal is narrower and more useful: to help you build an honest, evidence-based picture of what a role tends to pay so you can make informed decisions. Treat everything here as background information rather than financial advice, and remember that your own circumstances, location, and level matter more than any single chart.
There is also a practical reason to do this early. Engineering loops often run to five or six stages over several weeks. If the band turns out to be well below where you can credibly sit, you want to know that before you have spent four evenings preparing, not after a final-round offer you then have to unwind. Research is partly a filter for which processes are worth your time.
Good research does not produce a single magic number. It produces a defensible range and an honest confidence level. If your homework gives you one precise figure, you have probably anchored on a single data point rather than read the market.
Start with public, structured data
The most reliable starting points are sources that aggregate many data points and publish their method. They will not be exact for one company, but they give you a defensible middle.
A few worth knowing:
- Levels.fyi collects self-reported compensation for tech roles, broken down by company and level. It is strongest for larger US technology firms and weaker for small companies and non-US markets.
- Glassdoor salary pages cover a broad range of employers and countries, though the data is self-reported and can mix titles together.
- For US occupational figures, the BLS Occupational Employment and Wage Statistics is a primary government source with median and percentile wages by occupation and area.
- For UK context, the ONS labour market pages give macro wage trends, though they are not a company-level pay database.
Read at least two of these for any role, because each has a different bias. When they roughly agree, you have a usable range. When they diverge sharply, that gap is itself information, usually about level or location.
It helps to understand what each kind of source is actually measuring, because the biases are systematic rather than random.
| Source type | What it measures | Main bias | Best used for |
|---|---|---|---|
| Crowd-sourced (Levels.fyi, Glassdoor) | Self-reported individual packages | Skews toward people who earn more and want to share, and toward large tech firms | Company and level comparisons in tech |
| Government statistics (BLS, ONS) | Survey-weighted occupational wages | Broad occupation buckets, lags the market by a year or more | A sanity-check floor and macro trend |
| Live job adverts | Employer-stated good-faith ranges | Wide bands, sometimes spanning two levels | The current, local market for one title |
| Recruiter or agency benchmarks | Placements they have actually made | They have an incentive in the number | A reality check, treated with care |
The pattern to internalise is that crowd-sourced sites read high, government statistics read low and late, and live adverts sit in between but with wide bands. Triangulate across all three and the truth usually lives in the overlap.
Use the job adverts themselves
Pay transparency rules have made job adverts a much better data source than they used to be. Postings covered by New York's pay transparency law and New York City's salary transparency rules must show a good-faith range. The EU Pay Transparency Directive pushes employers toward giving pre-interview salary information ahead of its mid-2026 deadlines. The UK has moved more slowly, with government guidance encouraging transparency rather than mandating ranges. Primary sources are the New York State pay transparency law, the NYC salary transparency page, and the EU pay transparency policy.
In practice you can often find several current adverts for the same role, read the posted ranges, and build your own small sample. Look at the company you are interviewing with, then three or four direct competitors hiring for a similar title. The spread across those adverts is usually a better guide than any single database figure.
A few habits make advert research more accurate:
- Save the exact band and the date you saw it. Adverts get edited and pulled, and you want a record of the original number.
- Read the responsibilities, not just the title. A band attached to "owns a service end to end and mentors two engineers" is a senior band wearing a mid-level title.
- Note whether the range is base only or total. US adverts increasingly state base, while some European adverts quote a total figure that folds in bonus.
- Watch for bands that are suspiciously wide. A range like 70k to 130k usually means the advert covers two levels, and the company will slot you into one of them after they decide your level.
When you have collected five or six adverts for the same role, you can write them down as a quick distribution rather than a single guess.
Backend engineer, mid-level, London (adverts seen 2026-06)
- Company A (interviewing): £75k to £95k base
- Competitor B: £80k to £100k base
- Competitor C: £70k to £90k base + 10% bonus
- Competitor D: £85k base, equity unstated
- Government floor (ONS-ish): median well below the above
Read: clustered around £80k to £95k base for the level I am targeting.That small table is more persuasive than any single chart, because you assembled it from current, local, like-for-like postings and you can see the spread with your own eyes.
Read the number in context
A salary figure on its own means very little. Before you treat any range as relevant to you, line it up against a few things.
Level and scope
The same title can sit at very different levels. A "software engineer" range that looks low might be an early-career band, while a senior band under the same title sits far higher. Whenever a source lets you filter by level, do it, and try to match the level to the scope described in the job advert rather than the title alone.
Levelling is the single biggest source of error in salary research. Companies map titles onto internal ladders that rarely line up, so one firm's "Senior Engineer" is another's "Engineer II". The way through is to stop reading titles and start reading scope: what the role owns, who it influences, and how much ambiguity it absorbs.
| Signal in the advert | Likely level |
|---|---|
| Works on well-defined tasks with guidance | Early-career |
| Owns features end to end, limited oversight | Mid-level |
| Owns a service or domain, sets technical direction, mentors | Senior |
| Drives cross-team architecture, influences strategy | Staff and above |
Match the role you are looking at to the row that fits its responsibilities, then pull salary data for that level rather than for the title printed at the top of the advert.
Location and remote policy
Location changes everything. London, a regional UK city, a high-cost US metro, and a fully remote role priced on a single global band are four different markets. Some companies pay remote staff by their location, others use one band for everyone. A figure pulled from a San Francisco data set tells you almost nothing about a role based in Manchester.
When a role is remote, find out early which model the company uses, because it changes the number more than almost anything else:
- Location-adjusted pay ties your band to where you live, so a remote role can pay very differently for two equally skilled people.
- Single-band pay sets one number regardless of location, which tends to favour people outside the most expensive cities.
- Hub-anchored pay sets the band to a specific city even though the role is remote, so read the advert for which city it names.
Total package, not base alone
Base salary is one component. Bonuses, equity, pension contributions, and benefits can move the real value of a role significantly, and they vary by company stage. A startup may pair a lower base with equity that carries both upside and risk, while a larger firm may offer a higher base with restricted stock. When you compare ranges, try to compare like with like, and note where a figure is base only versus total compensation.
To compare offers and adverts honestly, break the package into its parts rather than reacting to the headline:
- Base salary, the guaranteed part that drives pension and mortgage maths.
- Bonus, and whether it is contractual, discretionary, or tied to company performance you cannot control.
- Equity, noting whether it is liquid public stock or illiquid private shares, and over what vesting schedule.
- Pension contribution, which in the UK is a meaningful and easily overlooked difference between employers.
- Benefits with real cash value, such as private healthcare, extra leave, or a learning budget.
A higher base with no bonus can beat a lower base with a large but discretionary one. The only way to see that is to write the components down side by side.
A worked example, start to finish
Theory is easier to trust with a run-through. Suppose you are a backend engineer with about four years of experience, based in London, with a first call booked for a "Backend Engineer" role at a mid-sized fintech.
Fix the level. The advert says the role owns a payments service end to end and mentors a junior engineer. That is mid-level edging into senior, so you pull data for the mid to senior band rather than the entry band.
Gather structured data. Levels.fyi and Glassdoor both put mid-level London backend base pay in the low to mid eighties, with senior pushing past a hundred. The ONS figures sit lower, as expected, since they cover all software occupations nationally rather than London fintech.
Gather live adverts. You find five current postings for similar roles and write them into the small table shown earlier. They cluster around £80k to £95k base, with one competitor adding a ten per cent bonus.
Reconcile. Structured data and adverts agree closely once you control for level and location, and that agreement is what gives you confidence. Equity at the target company is private and illiquid, so you treat it as upside rather than guaranteed value.
Write the one-line read. "Mid-to-senior backend, London: base clusters £80k to £95k, confidence medium-high, equity is private so discount it." When the recruiter opens with a band of £78k to £88k, you are not guessing. You can see it sits at the lower-middle of the market for the level described, and you can ask calmly where in the band this role is pitched and how level is decided.
That is the point of the exercise. You are not memorising a number to recite. You are building enough context that the recruiter's figure lands as one data point among several rather than as a verdict.
Distilled into a reusable loop, the routine is: fix the level from scope, pull two structured sources, collect five live adverts, normalise everything to base, then write one honest line with a confidence level. The last step is the one people skip, and it is the one that stops you quietly upgrading a shaky guess into a firm belief.
Build a one-page picture
Pull your research into a short note you can actually use. It does not need to be elaborate.
Role / level: backend engineer, mid-level, London
Public data midpoint: from Levels.fyi and Glassdoor
Advert ranges seen: company advert + 3 competitors
Base vs total: note which figures include bonus or equity
Confidence: high / medium / low, and whyThe confidence line matters. If you only found two stale data points, say so. A wide, uncertain range is honest and still useful, because it stops you from anchoring hard on a number you cannot defend.
When the public data thins out
Not every role sits on a rich pile of data. How far you can trust your research depends on how densely the market is measured, and that density varies far more by role type and region than by anything else. Read the density before you read the number.
Some roles are heavily sampled. Common software titles in large, transparent markets show up in thousands of self-reported packages and hundreds of live adverts, so a triangulated range lands with high confidence. Others are barely measured. A niche specialism, a brand-new job family, or a role in a market where pay is rarely disclosed may surface only a handful of stale figures, and no amount of averaging turns four data points into a reliable band.
| What you are researching | Typical data density | What to lean on |
|---|---|---|
| Common title, large transparent market | Rich crowd-sourced and advert data | Triangulate normally, hold a tight range |
| Common title, smaller or low-disclosure market | Thin crowd-sourced, some adverts | Weight live adverts, widen the range |
| Niche or brand-new role | Sparse everywhere | A couple of honest recruiter conversations, confidence stated as low |
Two forces widen the range whatever the role. The first is levelling: the same title can sit a full band apart between companies, so the thinner your data, the more you should reason from the scope the advert describes rather than the title printed on it. The second is package mix. As a role gets more specialised, bonus and equity tend to make up a larger share of the total, and two packages with the same base can differ enormously in real value, so you compare component by component rather than trusting a single headline.
The rule holds everywhere and for every region: the less public data you can find, the wider you hold your range and the lower the confidence you attach to it. A wide, honestly-labelled range beats a precise number assembled from too little.
Where salary research quietly goes wrong
A handful of errors account for most bad salary research. Watch for them in your own work.
| Mistake | Why it misleads | Better approach |
|---|---|---|
| Reading the title, not the scope | Titles do not map across companies | Level by responsibilities |
| Trusting a single data point | One report is an anecdote | Triangulate three or more |
| Mixing locations | A US figure tells you little about a UK role | Filter to the exact market |
| Comparing base to total | Inflates or deflates the gap | Normalise everything to base first |
| Using stale data | Pay moved fast in some areas | Prefer figures under a year old |
| Treating equity as cash | Private equity may never pay out | Discount illiquid equity heavily |
| Anchoring on a wish | Research becomes self-justification | Separate what you found from what you want |
Sense-check before you rely on it
Beyond the mistakes table, two habits keep your research grounded. Prefer recent data, since pay in some areas moved quickly and a figure from three years ago may mislead. And cross-check currencies and tax assumptions, because a headline US number is pre-tax and excludes the healthcare and pension differences a UK package usually bundles in.
It can help to picture the contrast. Weak research is one Glassdoor average, read off the title, with no level filter, treated as a hard target. Good research is a small distribution assembled from live adverts, cross-checked against two structured sources, normalised to base, levelled by scope, and summed up in one honest line with a confidence level. The first produces a number you cannot defend. The second produces a range you can explain and the calm that comes from not guessing.
Where this fits in your prep
Knowing the range early helps you decide whether to invest in a long process and lets you talk about money calmly when the topic comes up, without inventing competing offers or claiming a vague "market rate" you cannot back up. Treat the recruiter's stated range as another data point to reconcile with your own, not as a verdict, and treat your own figures the same way.
It also frees up attention for the parts of the interview you can control. Once the money question is grounded in evidence rather than nerves, you can put your preparation into the rounds that decide the outcome. Pair this research with a clear plan for answering the salary expectations question when it lands on the recruiter call, and keep negotiating your compensation as a separate step for after an offer arrives, so the salary conversation becomes one more well-prepared part of the process rather than a moment you improvise.
FAQ
Should I research salary before or after the first recruiter call? Before. The first call often includes the recruiter stating a band or asking for your expectations. If you have done the work, their number lands as a data point you can place in context rather than a figure you have to react to on the spot.
What if I can only find data for the US and the role is in the UK? Treat the US figure as a different market, not a translation. Use it only to understand company tier and level structure, then build your actual range from UK adverts and UK-weighted sources such as the ONS labour market pages.
How many data points are enough? There is no exact number, but a useful rule is two structured sources plus around five live adverts for the same role and location. Below that you can still form a view, but you should hold it loosely and say so in your confidence line.
The advert lists a huge range, like 70k to 130k. What do I do with that? Assume it spans two levels. Use the scope description to decide which level the role you are interviewing for actually sits at, and read the half of the band that matches.
Is it worth paying for premium salary data? Usually not for a single search. The free combination of crowd-sourced sites, government statistics, and live adverts covers most roles well. Paid data is more justifiable if you are researching a niche or senior role where public datasets thin out.
Sources
- Levels.fyi collects self-reported compensation by company and level, strongest for larger technology firms.
- ONS labour market statistics for UK macro wage trends and a national sanity-check.
- New York State pay transparency law sets the good-faith range requirement that made many adverts a usable data source.
- NYC salary transparency rules for the city-level disclosure requirement on job postings.