Salary surveys in the UAE: what to check before you trust one

Every UAE salary survey looks authoritative on the page: clean tables, confident ranges, a named year. The differences that matter are underneath, and most surveys do not volunteer them. Before a figure from any survey reaches an offer, a band, or a board paper, run these five checks. They apply to every dataset, including ours.

Check 1: sample size per row, not per survey

A survey covering “thousands of data points” can still have three of them behind the row you are about to use. Aggregate sample size is a marketing number; the row-level count is the one that decides whether your figure is a benchmark or an anecdote.

Ask for the count behind the specific role, level, and market you care about. If the survey cannot break its sample down to the row, treat every row as thin. This is why the Tenure Pay Index prints the source count next to each band rather than quoting one big number for the dataset: a band backed by 24 sources and a band backed by 4 should not be read with the same confidence, and pretending otherwise helps nobody.

Check 2: the date of the data, not the date on the cover

A survey published in January 2026 typically collected its data in mid-to-late 2025. The cover year describes the print run, not the market. Ask when the underlying observations were gathered and when the specific row was last refreshed. If the answer is “annually”, every figure is up to a year stale on the day it is newest, and older by the time you use it.

Refresh cadence is a structural property, not a detail. The Pay Index refreshes quarterly and stamps every band with its refresh date, so the age of a figure is visible before you rely on it, and a stale row cannot hide behind a current cover.

Check 3: what “verified” actually means

“Verified data” is doing heavy lifting in most survey marketing. Get the operational definition: verified by whom, against what? Self-reported figures with a plausibility filter are not the same evidence as offer letters. Scraped job adverts, which state advertised ranges rather than agreed pay, are weaker still.

Tenure’s definition, published in full on the methodology page: every band is built from primary sources, meaning extracts from established Gulf compensation surveys and anonymised offer-letter data, and every row is checked by a professional before it enters the dataset. Verified sources weigh more than unverified ones, recent observations weigh more than older ones, and outliers are excluded before a band is computed. Whatever dataset you use, the answer to “what does verified mean” should be that specific.

Check 4: whether tiers and cities are blended

UAE pay for the same title splits sharply by employer tier, and Dubai and Abu Dhabi diverge at the senior bench. A survey that publishes one national, all-tier range per title has averaged those differences away, and the resulting figure competes against the wrong benchmark on every offer: too high for a regional employer, too low for an international one.

Check the survey’s segmentation. If it cannot tell you which tier and market a range describes, the range describes a blend that no single employer will ever pay.

Check 5: percentiles, not just a minimum and maximum

A min-max range invites reading the midpoint, and the midpoint of a wide range is not a market rate. What you need is the shape of the distribution: P10, P25, median, P75, P90. That is what turns “the range is 30 to 70” into “this offer sits at P35”, which is a sentence you can act on. If a survey publishes no percentiles, it is publishing a boundary, not a distribution.

Why the honest answer is sometimes “no data”

The test that separates a serious dataset from a padded one is what happens when the evidence is thin. Tenure publishes a band only once at least three independent sources support it; below that line the band is held back entirely, which is why some roles you look for are not there yet. The same principle sets the market coverage: UAE and Saudi only, because those are the two markets where the verified source pool per sector is deep enough to publish credible distributions. A survey with a figure for everything, everywhere, has answered the coverage question with padding.

The five checks reduce to one habit: ask where the number came from, and do not accept an answer that describes the whole survey instead of the row. For the working method that uses these checks on your own team, see salary benchmarking in the UAE, or run the free benchmark to see the evidence trail on a live band, source count included.