Pay-Transparency Reporting Is Due, but Three Payroll Systems Disagree: Is This a Data Project?
Reconcile worker population, pay components, hourly and annual measures, and worker categories before treating an EU pay-transparency reporting gap as a software project.
Signals to watch
- A named legal employer and reporting year have conflicting metric outputs from payroll or HR systems
- Worker population, pay definition, hourly/annual denominator and category mapping can be tested against the applicable national rule
- A reporting submission, worker-representative review or remediation decision has an owner and date
Three different gender-pay-gap figures do not automatically justify a new HR platform. The discussion becomes an EU pay-transparency data project when one legal employer and reporting year repeatedly produce irreconcilable results because worker population, pay components, hourly/annual measures or worker categories cannot be governed across source systems. A national-law question or one payroll correction should be routed first.
This is the problem an HR-data platform BD lead sees in authorized payroll, human-resources information system (HRIS) and employment-compliance Telegram groups. The useful Signal is a reproducible metric conflict attached to a reporting or worker-representative decision. Seeing it a day late can miss a data-definition workshop. Selling too early can turn a legal interpretation into a software migration.
Illustrative industry case — this composite scenario explains a decision pattern. It is not a customer story, employee record or commercial result.
One group fragment says:
“Payroll gap is 4.7%, HR dashboard says 6.1%, finance has 5.4%. First report scope is this year. Need one source of truth.”
The percentages are invented only to illustrate disagreement; they are not market data. The thread does not identify the legal employer, country law, worker count, reporting year, currencies, base/variable pay, hours, joiners/leavers, worker categories, data owners or calculation methods.
The first postmortem question is “same employer, same people?”
Directive (EU) 2023/970 required Member States to transpose the pay-transparency rules by 7 June 2026. Reporting is organization-specific, but national law determines execution. A multinational group dashboard can combine workers who do not belong to the same reporting employer.
Extract the legal employer, reporting country and previous-calendar-year population. Then reconcile employee identifiers, employment dates, contracts, leaves and exclusions under the applicable law. A headcount snapshot and a population of everyone paid during the year are not the same denominator.
If the three systems start with different people, do not compare their percentages yet. Produce a population difference file first: included only in payroll, only in HRIS, only in finance, and the rule or defect behind each difference.
The second question is “same definition of pay?”
The Directive defines pay as ordinary basic or minimum wage or salary plus other cash or in-kind consideration received directly or indirectly in respect of employment. Complementary or variable components can include bonuses, overtime compensation, allowances and other benefits.
Payroll may store actual paid amounts, HR may store contractual base salary and finance may aggregate ledger accounts. A bonus paid in March for the previous year’s performance can land in different periods depending on the extraction rule. Currency conversion and gross-versus-net fields can create further mismatches.
Create a component dictionary with source field, legal pay category, inclusion rule, reporting period, currency rule and owner. A final “total compensation” field without that dictionary is not auditable.
The third question is “same measure and denominator?”
Article 3 defines pay level using gross annual pay and corresponding gross hourly pay. Article 9 requires several metrics, not one generic gap: average and median gender pay gaps, gaps in complementary or variable components, the proportion receiving those components, distribution across quartile pay bands, and category-level gaps broken down by basic and variable components.
One system can calculate an unweighted average of annual salary; another can convert to hourly pay using contractual hours; a third can use actual paid hours. All three can execute their formulas correctly and answer different questions.
Record the metric name, numerator, denominator, unit, rounding, treatment of part-time work and period. Then recompute one metric from row-level source data. The first unexplained difference is the defect boundary.
Worker categories are a legal and data decision
The Directive defines categories of workers as workers performing the same work or work of equal value, grouped non-arbitrarily using objective gender-neutral criteria. A job title or grade alone may not represent the required category, while a machine-generated cluster cannot make the legal decision.
Preserve the category method, criteria, approval owner and version. Then map each worker with an effective date. If category definitions change after a worker-representative review, historical results need a controlled restatement rather than an overwritten column.
The 5% threshold is not an automated verdict
A joint pay assessment is required when three conditions all hold: reporting shows an average difference of at least 5% between women and men in a worker category; the employer has not justified it with objective gender-neutral criteria; and the unjustified difference has not been remedied within six months after reporting.
A dashboard can flag the numeric condition. It cannot decide whether an explanation is objective, gender-neutral and legally sufficient. Nor can a group message establish discrimination. Keep legal review and worker-representative participation visible in the workflow.
When this becomes a data project
Route the work by the first failure:
- legal interpretation: employer scope, national reporting rule or category criteria are unsettled;
- payroll correction: one source field or population defect can be corrected and controlled locally;
- recurring data remediation: several systems repeatedly disagree on identity, component, period, hours or category and no governed lineage can reproduce the required metrics;
- reporting operations: data is reconciled, but submission, publication or worker-representative workflow needs ownership.
The acceptance test for a data project is not “all dashboards show the same number.” It is: one approved worker population and component dictionary reproduce every required metric from traceable records, exceptions are visible, and legal/category decisions have named owners.
The DORA register article shows a different regulated relationship-record problem. The regulation-driven demand test prevents a deadline from becoming a project by itself; confidence scoring explains why a ranked message still needs evidence.
TOP Prospect can connect incomplete fragments from groups the user deliberately connects and may access, preserve source and time, merge obvious duplicates and rank the candidate for human review. It cannot access employee data, calculate a legally reportable metric, decide discrimination, contact workers or file a report. The pricing page describes the discovery boundary.
The three numbers become a software candidate only after the postmortem identifies why they disagree and shows that the same uncontrolled break will recur. Until then, the right deliverable may be a legal definition note or one corrected payroll extract.
Frequently asked questions
When do employers with 250 or more workers first report under the Directive?
Directive (EU) 2023/970 sets 7 June 2027 for the first report on the previous calendar year and annual reporting thereafter, subject to national transposition and implementation.
What are the first reporting dates for smaller covered employers?
Employers with 150–249 workers report by 7 June 2027 and every three years; employers with 100–149 report by 7 June 2031 and every three years. Member States may require reporting below 100 workers.
Does pay mean only base salary?
No. The Directive defines pay as ordinary basic or minimum wage or salary plus other cash or in-kind consideration received directly or indirectly, including complementary or variable components.
Does a 5% gap automatically prove discrimination?
No. A joint pay assessment is triggered when the report shows at least a 5% average gap in a worker category, the employer has not justified it with objective gender-neutral criteria, and it has not remedied the unjustified difference within six months.
Sources and further reading
How a Signal worth attention is found
See how Top Prospect finds and organizes Signals worth checking, keeps the original Telegram context, removes duplicates, and helps you decide what to review first. You decide whether to follow up and what to do next.
