FDE pay: why salary sites and job ads give different numbers

In its job ad for a Forward Deployed Software Engineer working with the US government, Palantir lists an estimated salary range of $135,000–200,000 a year. Meanwhile, the US page on Levels.fyi gives the median for a Forward Deployed Engineer as $220,000, above the top of Palantir’s range.

Neither is lying. Each answers a different question, and readers tend to collapse them into one: “How much does an FDE earn?”

For anyone working as an FDE or about to become one, the confusion has a price. Quote the wrong number when a recruiter asks about your expectations and you either undersell yourself or get ruled out for asking for a figure the company has never put in its base range. Before deciding which number to trust, you need to know what each one measures.

Each source measures something different

Levels.fyi is explicit: the figure it reports is total compensation, meaning base salary plus the annualised value of equity plus bonus. The data comes from salary reports that users submit to the site, not from pay that companies post.

Glassdoor measures something narrower. The methodology Glassdoor publishes for its Local Pay Reports refers only to median base salary for full-time workers in the US, calculated nationally and by metro area. If the FDE figure you see on Glassdoor is calculated on the same basis, equity and bonus are not in it.

A third view comes from job ads. Recruiting from Scratch analysed 924 FDE job postings and states plainly that its data comes from real job ads, not surveys. The result is a median of $195,000 “across all locations”, but it does not say which countries those locations are in.

That is the pay companies publish when they advertise a role, not the final offer. Exponent’s aced.io takes the same approach and cites a median base salary of about $190,000 across 135 US job postings, below the Levels.fyi median for total compensation.

Source	What it measures	Where the data comes from	Sample size	Published figure
Levels.fyi	Total compensation: base + equity per year + bonus	Salary reports submitted by users	Not given for the US page; UK 33, India 11	US: median $220,000 (as of 28 October 2026). UK: £140,302
Glassdoor	Median base salary (per the Local Pay Reports methodology)	Full-time workers in the US	Not given	Used here only to explain the method; no FDE figure to set alongside
Recruiting from Scratch	Posted pay	Real job ads	924 postings	Median $195,000, “all locations”, countries not stated
aced.io	Posted base salary	US job ads	135 postings	Median about $190,000
Palantir job ad	Estimated base range	The company itself, one US role	One role	$135,000–200,000

The gap is equity and bonus

Read Palantir’s ad closely and you find a line many people skim past. The $135,000–200,000 range is only the estimated base salary, and total compensation for the role may also include RSUs and a sign-on bonus.

The $220,000 Levels.fyi median exceeds that $200,000 ceiling. The most plausible explanation is that the difference lies in equity and bonus, exactly the items Palantir leaves out of its posted range.

Take a hypothetical offer with a base of $170,000 and a $200,000 equity grant vesting evenly over four years. By the Levels.fyi method, total compensation is $170,000 + $50,000 = $220,000 a year. A site that measures only base pay would show $170,000: the same person, the same offer, and two figures $50,000 apart.

At some companies the gap is far wider. According to aced.io, at OpenAI (with PPUs) and Anthropic (with RSU-heavy packages), equity can roughly double the cash component. In that case a site that measures only base pay may show you less than half the real value of the offer.

The 2026 FDE salary report from getperspective.ai, which compiles 1,200 data points, makes the same point: base salary is the floor, not the ceiling.

How much can 33 reports tell you?

Sample size is the second trap. The UK page on Levels.fyi gives a median of £140,302 for FDEs, but it rests on only 33 salary reports.

With 33 values, the median is the 17th. Add a few new reports that are well above or below the rest and a different person sits in 17th place, and the figure on the page changes with them.

The India page is thinner still: 11 reports. At that size, a handful of offers from a single company paying very high or very low is enough to skew the whole median. So each time you quote a Levels.fyi figure, note the sample size and the date you checked it.

Don’t count one source twice

The third trap is harder to spot: aggregate reports reuse the very sources you have already read. getperspective.ai says it compiled data from five public sources between January and May 2026, including Glassdoor and Levels.fyi.

If that report’s figure matches Levels.fyi, it is not independent confirmation, because part of the data comes from the same place. When you gather figures, trace where each one originates before treating it as a separate “vote”.

Country is the next source of distortion. getperspective.ai uses only US data and says international pay ranges are just 50–70% of US levels. The $195,000 median from Recruiting from Scratch, by contrast, covers “all locations”, so you cannot tell how much of it comes from postings outside the US.

So don’t set an “all locations” figure beside a US-only figure and conclude that the market is rising or falling. Nor should you convert pounds or rupees into dollars yourself and compare the result with the US page: look for a source in the same country and currency as the market you are applying to.

So which number should you trust?

Trust the number that answers your question. If you need to know how much cash the company will pay each month, the best benchmark is the base range in that company’s own job ad, read together with the line listing what falls outside the range. Use Glassdoor only as a cross-check on base pay.

If you need to compare two full offers, or want to know how much room there is to negotiate on equity, use Levels.fyi on three conditions: check the sample size, note the date you looked, and make sure the page you are reading covers the country you are targeting.

Aggregated job-ad data tells you what the market is offering, not what people actually take home.

Three habits work in any market. When reading a job description, separate the base range from everything outside it.

When a recruiter asks about your expectations, answer in terms of total compensation and ask for the breakdown: base, equity per year, bonus and sign-on bonus. When you have an offer in hand, compare it with a source that measures the same thing in the same country, not with a median of unknown origin.

This is also what FDEs do every day at client sites: before trusting a number, ask how it was measured, from what sample and when. Anyone used to asking those questions of a client’s data has no reason to take a salary site’s figure on trust.