Your Engineering Job Ladder Is Now a Legal Document: The Pay Transparency Deadline Almost Nobody Prepared For

What Actually Changed in June, and Why Your Legal Team Said Nothing

On 7 June 2026, the deadline for EU member states to transpose the Pay Transparency Directive into national law passed. If nothing happened in your engineering organisation that week, no all-hands, no compensation memo, no frantic Slack thread about job ads, you are in the overwhelming majority. That is exactly what makes this one dangerous.

The reason for the silence is that most member states missed the date. According to Lewis Silkin's June 2026 FAQs on the directive, only four member states met the deadline: Italy, Lithuania, Malta and Slovakia. Morgan Lewis tracked the delayed group, with the Netherlands, Sweden, Czechia and Denmark targeting 1 January 2027, and Sweden's implementation paused pending discussion at EU level.

So when your general counsel says "not yet, in our jurisdiction", they are almost certainly right. Before national transposition, private employers are not required to comply, and employees cannot bring claims for non-compliance with the directive itself. That is a correct legal answer to the wrong question. The question is not "when must we comply", it is "when does the data we will be judged on start accumulating", and the answer to that one is different.

Under the schedule set out in the directive, the first gender pay gap reports fall due on 7 June 2027 for employers with 250 or more workers, then annually; employers with 150 to 249 workers report from the same date every three years; and the 100 to 149 band starts in 2031, also every three years, as Figures sets out in its employer guide. Those 2027 reports are built on 2026 payroll data. The measurement year is the one you are living in right now, which means the numbers you will publish are being created by offers your hiring managers are signing this quarter.

This has the exact shape of GDPR in 2016. Everyone treated it as a legal date until they discovered it was a data-model date, and the organisations that started twelve months early spent the deadline calmly while everyone else rewrote schemas in a panic. The difference this time is that the data model in question is your job ladder.

Why a Pay Directive Lands on the Engineering Org Chart

The instinct is to file this under HR. That instinct is wrong, and the reason is the directive's unit of analysis. It does not compare individuals to individuals. It compares "categories of workers", meaning groups of people doing the same work or work of equal value, grouped in a non-arbitrary way on objective, gender-neutral criteria.

"Work of equal value" has a specific legal meaning that has almost nothing to do with how engineering organisations normally group people. As WTW explains in its guidance on categorising workers, value is assessed on four factors: skill, effort, responsibility and working conditions, independent of the actual tasks and duties. If a senior backend engineer, a senior QA automation engineer and a senior data engineer come out equivalent on those four dimensions, they are plausibly one category, and any systematic pay difference between them needs a documented, gender-neutral justification. Most engineering organisations pay those three roles differently and have never written down why in terms that would survive a tribunal.

The common shortcuts do not work either. Syndio's analysis of job architecture requirements is blunt about it: grouping by job title, by level within a job family, by function or department, or by simply adopting your existing pay grades does not by default establish equal value, because those structures were built for other purposes. Engineering ladders in particular were designed to manage promotion conversations and headcount budgets. They were not designed to be read adversarially by somebody trying to prove that two groups doing comparable work were paid differently.

And they will be read adversarially, because the burden of proof moves. Lewis Silkin's reading of the directive is that non-compliance shifts the burden of proof onto the employer in an equal pay claim, unless the breach was manifestly unintentional and minor. In practice that inverts the default: it is no longer for a claimant to demonstrate discrimination, it is for you to demonstrate that a pay difference is explained by objective criteria you can evidence. Your levelling rubric stops being an internal management tool and becomes the primary exhibit.

That is why this is an engineering leadership problem. Nobody in HR can write down what distinguishes a Senior from a Staff engineer in a way that holds up, because nobody in HR makes that call. You do.

The Job Ad Problem: Publishing a Band You Can Defend

The recruitment obligations are the part that bites first, because they apply to employers of every size regardless of headcount thresholds, and because they change something your hiring managers do every week.

Two rules matter. First, candidates must be given the initial pay level or its range early enough to allow an informed negotiation, which in practice means it goes in the advert. Second, employers must stop asking applicants about their current or past pay. Lewis Silkin's guidance also warns that overly broad ranges risk non-compliance, which closes the obvious escape hatch. Advertising an engineering role at "45,000 to 110,000 euros depending on experience" is not a pay range, it is a refusal to publish one, and it will be treated as such.

This creates an internal problem before it creates an external one. The moment you publish a real band for a Senior Backend Engineer, every Senior Backend Engineer already on your team can compare it to their own payslip. Pay compression that was survivable while invisible becomes a retention event. And they have a mechanism to pursue it: workers can request their own pay level plus the average pay by gender for colleagues doing equivalent work, along with the objective criteria used for pay decisions, and Figures notes the employer must respond in two months and remind staff of the right annually.

For engineering specifically, the hardest collision is with specialist premiums. If you are paying twenty per cent above band for machine learning and platform engineers because that is what the market costs, you now have to decide which is true: either those roles are a different category of workers, distinguishable on documented differences in skill and responsibility, or they are the same category and the premium is an unjustified differential. Both answers are available. Only one of them can be true for any given pair of roles, and you have to pick before somebody picks for you. This is the same underlying discipline that rebuilding a hiring loop around evidence rather than vibes demands, applied to compensation rather than assessment.

One practical note that engineering leaders consistently underrate: the salary history ban removes the crutch that made sloppy levelling survivable. Anchoring an offer on what a candidate earned previously is how pay inequity propagates between employers, and it is also how organisations avoided ever deciding what a level is actually worth. Take that away and you are forced to have a number that comes from the ladder. If the ladder cannot produce one, the ladder is the thing that is broken.

Gender Pay Gap Reporting Is a Data Engineering Project

Strip away the legal framing and the reporting obligation is a query. You need the gap between men and women in average pay, the gap in complementary and variable components, the proportion of each gender in each quartile pay band, and the gap within each category of workers. That last one is the expensive dimension, because it requires your category definitions to be stable, documented and reproducible a year later.

The blocker is almost never the statistics. It is that the inputs live in four systems that were never joined. Base pay sits in country payroll instances that differ per legal entity. Variable compensation sits in a bonus tool or a spreadsheet owned by finance. Equity sits in a cap table platform. Hours, FTE fractions, leave and start dates sit in the HRIS, and the employee-versus-contractor flag in that HRIS is, in most companies, wrong for a non-trivial slice of the population.

The readiness data reflects this. Aon's 2026 Pay Transparency Pulse Survey of more than 1,000 HR professionals found that 42% name data quality as their primary compliance challenge, and Aon's earlier reading put the share of companies that felt ready for full pay transparency at 19%. More than three quarters had already implemented some pay transparency measure. Implementation and readiness turned out to be different things, and the gap between them is a data pipeline.

Framed honestly, this is a warehouse and transformation project with a legal deadline attached: entity-level extracts, a canonical person identity across systems, FTE normalisation, currency and period alignment, a versioned mapping from internal job codes to worker categories, and reproducible snapshots so that next year's joint pay assessment can re-run this year's numbers and get the same answer. Anyone who has built a regulatory reporting pipeline in fintech will recognise the shape immediately.

It is also work that HR cannot staff, which is how it lands on the engineering roadmap in Q4 with no budget line. If your in-house team is already committed to the product roadmap, this is a well-bounded, high-leverage piece to hand to an external team: the requirements are externally specified, the surface area is contained, and the deliverable is testable. It is precisely the kind of engagement we run with dedicated data engineers alongside a client's existing platform team, without pulling product engineers off the roadmap to write payroll joins.

What the Numbers Say About Engineering Specifically

European tech has a specific shape of pay gap, and understanding it changes what you should worry about. Ravio's 2026 analysis of European tech compensation puts the unadjusted gender pay gap at 23% and the adjusted gap, the portion left unexplained once job-related factors are accounted for, at 2.4%.

The distance between those two numbers is the whole story. A 2.4% adjusted gap says that paying two people differently for the same job is not the dominant mechanism. A 23% unadjusted gap says that men and women are distributed very differently across levels and functions. The driver is representation, and Ravio's accompanying work on women in European tech puts women at 21% of executive positions across the industry.

The pipeline numbers underneath are starker. Eurostat's statistical overview of ICT education records that in 2025, men accounted for 83.4% of the 3.3 million people employed in the EU with an ICT education. Across the broader scientist and engineer population, Eurostat reported in 2026 that women made up 40.8% in 2025 while remaining a majority of science and technology employment overall.

There is one figure that should genuinely alarm hiring managers, though, because it is created at the moment of the offer rather than inherited from the pipeline. Ravio's data shows a male software engineering professional receiving a new hire salary 3.3% higher than an equivalent female candidate. That is a gap manufactured during negotiation, in a process you control, in the exact place the directive is aiming: publish the range, do not ask about salary history, make the number come from the ladder.

The strategic consequence is uncomfortable and worth saying plainly. If your company is engineering-heavy, your published unadjusted quartile figures will look bad even if your same-role pay is clean, because your quartiles reflect a talent pipeline that Eurostat has been documenting for a decade. You will be explaining a large public number that is mostly about representation. The organisations that come out of this well will be the ones who publish the unadjusted number alongside the adjusted one, explain the difference credibly, and can point at what they are doing about representation. The ones who come out badly will be the ones who see the headline figure for the first time three weeks before the filing date.

The Threshold That Triggers a Joint Pay Assessment

The mechanism that turns reporting into obligation is the joint pay assessment, and it has three conditions, set out in Lewis Silkin's FAQs: a gap of 5% or more exists in any category of workers, it cannot be objectively justified on gender-neutral grounds, and the employer has not corrected it in the six months following the report. Meet all three and you must run a formal assessment jointly with worker representatives.

Note the phrase "any category of workers". This is not a company-level test. A clean overall number protects you from nothing if one category, say your platform engineering group or your QA function, is out by six per cent. The more granular your categories, the more independent chances you have to trip the threshold, which creates a genuine tension: granular categories are easier to justify on equal-value grounds but produce more tests, while broad categories produce fewer tests but are harder to defend as genuinely comparable work.

How likely is this in practice? Figures reports that 82% of companies joining its platform have pay gaps exceeding the 5% threshold somewhere in their workforce. Treat a joint pay assessment as the base case rather than the exception, and the planning question changes from "how do we avoid this" to "what will we be able to show when it happens".

What you can actually control comes down to two things, and neither is a trick. The first is category design with documented, gender-neutral criteria written before you see the numbers, because criteria authored after the gap is known are worth very little. The second is remediation budget. A gap that is real has to be closed with money, and the difference between an organisation that handles this well and one that does not is usually whether somebody put a line in next year's compensation plan for it in advance.

Where Outsourced and Nearshore Teams Sit in the Scope Map

Almost every engineering organisation of meaningful size now runs a mixed population: employees, contractors, agency staff, employer-of-record arrangements and outsourced teams. Working out who lands in the reportable population is a real question, and it is worth answering precisely rather than assuming.

The baseline is that the directive covers workers with an employment contract or employment relationship as defined by each member state. Gibson Dunn's analysis for companies with EU-based employees notes that non-EU employers are pulled in where they have employees based in the EU. A supplier headquartered outside the EU is not in scope simply because its client is European; it is in scope where it employs people inside the EU. CXC's guide to the contingent workforce dimension works through the rest: agency workers and employer-of-record workers usually sit in scope via their legal employer, while genuinely self-employed contractors generally do not.

That last word, genuinely, is doing a lot of work, and this is where I want to be direct rather than clever. The scope boundary is not an optimisation opportunity. Restructuring employment relationships to shrink the population you have to report on is exactly the substance-over-form manoeuvre the framework anticipates: Ravio's employer guide points out that the existence of an employment relationship is determined by the facts of the working arrangement and not by the label on the contract. The misclassification exposure you would create is larger and nastier than the reporting obligation you would avoid, and it stacks on top of the tax and permanent establishment problems covered in our piece on hiring developers abroad without creating a misclassification problem. Do not go there.

The operationally useful insight from CXC's guide is different and more subtle: the supplier may be the legal employer while the client still controls the requisition, the approved range, the interview process and the workforce data. Responsibilities can be allocated by contract, but statutory liability does not transfer with them. So the work is a mapping exercise, done per engagement rather than per vendor. For each one, write down four names: who is the legal employer, who controls recruitment and the pay range, who holds the data, and who answers if a worker or a regulator asks a question. Then make sure your supplier contracts actually oblige cooperation with audits, notification of changes, and preservation of recruitment evidence such as approved ranges and screening records. A policy is not an audit trail.

This is one area where a senior-led nearshore partner is genuinely simpler than the alternatives, and the reason is structural rather than promotional. StepTo employs its engineers directly in Serbia rather than assembling a chain of freelancers behind an account manager, so there is one legal employer, one pay structure, and one party that can answer questions about how a rate was set. Our rate card is published, which means the commercial rate behind every role is a documented number rather than a negotiated mystery, and it maps cleanly onto a client's own levelling framework when their compensation team asks how our Senior corresponds to theirs. If you are weighing this against a marketplace of individual contractors, the difference between a dedicated team and a project-based arrangement is exactly the difference between one accountable employer and a scope map with gaps in it.

A Ninety-Day Plan for Engineering Leaders

The temptation is to wait for your jurisdiction's transposing law and then react. Given that the 2026 payroll year is the one being measured, and given that job architecture work takes two to three quarters in any organisation large enough to be in scope, that is a plan to be late. Here is what actually fits in a quarter.

In the first month, do the inventory. Pull every engineering role currently in your HRIS and map it to your published ladder. In most organisations this exercise alone surfaces something awkward: titles that exist in payroll but not in the ladder, people carrying a level that no longer matches what they do, and whole functions such as QA, data or SRE that were never levelled against the core engineering track at all. You cannot categorise workers you cannot enumerate.

In the second month, write the criteria before you look at any gap numbers. For each level and each specialism, document what distinguishes it in terms of skill, effort, responsibility and working conditions, the four factors the directive uses, and state explicitly which specialisms you consider equal value to each other and why. This is the document that carries the burden of proof later. Criteria written before the data are credible; criteria written afterwards read like a justification, because that is what they are.

In the third month, build the pipeline and run the numbers privately. Join payroll, HRIS, variable compensation and equity into a single reproducible dataset, compute the gaps by quartile and by category, and find out where you sit relative to the 5% threshold while you still have quarters rather than weeks to respond. Then take the two or three worst categories to whoever owns the compensation budget with a costed remediation plan. That conversation goes very differently in 2026 than it will in 2027.

Running that in parallel with a product roadmap is the actual constraint, which is the argument for adding capacity rather than resequencing the roadmap. The pipeline work is well-specified and self-contained, and the job architecture work needs senior engineering judgement that cannot be outsourced but can be facilitated. Our delivery process is built around exactly this split, and a dedicated development team from Belgrade sits in European working hours, which matters more than usual when the people you need in the room are your own compensation and HR leads rather than an engineering counterpart.