The deadline for transposing the EU Pay Transparency Directive (Directive (EU) 2023/970) was 7 June 2026. It has passed, and the compliance picture is a mess. That mess is now your problem, because the first thing the Directive does is force you to publish a number, and the sources you would use to set that number disagree with each other by more than 40%.
Start with who actually met the deadline, because even this is contested. Remote Work Europe reported that only Slovakia and Lithuania had passed full legislation. DCI Consulting counted four: Slovakia, Italy, Lithuania and Malta. The gap is definitional, not factual. Malta and Poland transposed only the pay transparency articles and left the reporting regime alone, so whether you count them depends on whether partial counts. Remote Work Europe also published on 27 May 2026, before the deadline, so its figure was a forecast. Treat any single transposition tracker as provisional.
Meanwhile the Netherlands, Denmark and the Czech Republic are openly targeting January 2027, and Sweden halted its transposition on 26 March 2026 while it seeks to renegotiate.
Why this lands on UK desks
The UK is not in scope as a jurisdiction. Your EU entities are. Any UK employer with subsidiaries, branches or EU-based remote employees must comply for those headcounts. If you run engineering out of London with a Dublin, Berlin or Warsaw pod, you will be publishing ranges for part of your team and not the rest.
That asymmetry does not survive contact with engineers. Post a band for a Warsaw backend role and your London backend engineers will find it the same afternoon.
The domestic timeline is closing anyway. The Cabinet Office launched its Equal Pay and Pay Discrimination consultation in July 2026, closing 27 October 2026, proposing a statutory duty to publish pay in job adverts or in writing before interview. Lewis Silkin notes the UK proposal is narrower than the Directive: there is no salary history ban in it, and no right for workers to request comparator pay. A government response is unlikely before early 2027.
Three sources, three different UK medians
Here is the problem in one table. All three of these purport to describe the same job in the same country.
| Source | Basis | UK median, software engineer | Data as at |
|---|---|---|---|
| IT Jobs Watch | Advertised salaries, 3,836 quotes | £70,000 | 6 months to 11 Aug 2026 |
| Ravio | HRIS payroll data | £69,600 (mid-level) | 2026 benchmarks |
| ONS ASHE 2025, Table 14 | Full-time employee survey | £56,914 | 2025 provisional |
The advertised and payroll figures agree almost exactly. The official statistic is £56,914, roughly 18% lower. That is not an error in the ONS data. ASHE covers the SOC group "programmers and software development professionals" across every sector and region, including local government, insurance back offices and manufacturing, and its annual figures normally cover people in the same job for over a year, so it excludes the new-hire premium entirely. It is a good measure of what the national profession is paid and a bad input for a London product engineering band.
Now the London figures, where the two commercial sources come apart badly.
| Source | London median | 25th percentile | 75th percentile |
|---|---|---|---|
| IT Jobs Watch | £100,000 | £77,500 | not published |
| Ravio | £72,000 (mid-level) | £65,000 | £84,000 |
A £28,000 gap on the same title in the same city. Two reasons, and both matter for how you use them.
First, sample and mix. IT Jobs Watch London rests on 740 salary quotes, and London job ads carrying the words "software engineer" skew heavily towards trading firms and financial services, which advertise ceilings. Ravio pulls from live HRIS integrations across 1,600+ companies, weighted towards venture-backed technology firms, and reports what is actually in payroll.
Second, advertised is not paid. An advertised range is a negotiating position. Payroll is the outcome.
Be sceptical of the trend lines too. The IT Jobs Watch UK median moved £60,000 in 2024, £80,000 in 2025, then £70,000 in 2026: down 14.3%, up 33.3%, down 12.5%. No labour market does that. Vacancy volume over the same window went 6,699, then 2,509, then 5,070. When the denominator swings by a factor of nearly three, you are reading changes in which employers happened to be hiring, not changes in the rate for the job.
For anchoring the bottom and top of a UK structure: graduate software developer advertised median is £25,000, the UK 90th percentile is £100,000, and Ravio puts senior software engineer at £110,200. Ravio also reports a median new-hire equity grant of 8.6% of base salary, which is the part of the package the Directive will not standardise for you.
The US comparison, and the same trap
| Source | Basis | US median | 25th | 75th |
|---|---|---|---|---|
| BLS OEWS, May 2025 | Base wage, all employers | $135,980 | $105,210 | $171,980 |
| Levels.fyi | Self-reported total comp | $195,000 | $137,000 | $280,000 |
A $59,000 gap, and again both are correct. BLS measures base wages across every US employer including defence contractors and regional banks. Levels.fyi measures total compensation, including equity, self-reported by a population that skews hard towards large technology companies. Comparing a BLS median to a Levels.fyi median is comparing base pay in Ohio to total comp at Meta.
The US is worth watching for a second reason: it has run this experiment already. Eighteen states plus Washington DC now have pay transparency laws, though counts of 16 plus DC also circulate, the difference being whether you count states that only require disclosure on request. California's SB 642 took effect on 1 January 2026.
The economics are not what most boards assume. Cullen and Pakzad-Hurson, in Econometrica in 2023, found that US state transparency laws led wages to decline by roughly 2%. The mechanism: transparency strips individual bargaining power, because an employer can credibly refuse a high offer to one person rather than trigger renegotiation with everyone else on the band. Transparency compresses. It does not inflate.
What to do if you are setting a band this quarter
Set the band from payroll data for your actual comparator set, then sanity check it against advertised data, and never the reverse. Advertised medians tell you what your competitors are willing to say out loud; payroll medians tell you what they pay. If you publish a range anchored to advertised London figures, you will have committed in public to a ceiling you cannot fund, and under the Directive's worker information rights your existing team will be entitled to ask why they sit below it.
One practical step before you publish anything: run your current engineers against the band you are about to advertise, and find every person already in post who falls below its midpoint. Those are your renegotiations, and they will arrive within a week of the first advert going live. Budget for them now, or narrow the band.
If you want your engineering bands pressure-tested against payroll data rather than job adverts before you commit them to a public advert, talk to Synerjy.