← All insights
Guide6 min read

The regulatory reporting process: from reconciling figures to proving outcomes

Most firms can show that their figures agree. Far fewer can show they are right.

The regulatory reporting process is the chain of steps that turns a firm's source data into the returns it submits to supervisors: extraction from source systems, mapping, the transformations and rules in code, manual adjustments, review and submission. For banks that means returns such as FINREP and COREP; for insurers, Solvency II quantitative reporting templates and IFRS 17 disclosures. Supervisors now expect firms to show how each reported outcome was produced. Most firms can show that their figures agree. Far fewer can show they are right.

What supervisors expect from regulatory reporting

Regulatory returns feed capital, liquidity and solvency decisions, so supervisors treat their accuracy as a matter of governance, not administration. In the UK, senior managers carry personal accountability under the senior managers regime, and from 2026 the boards of companies reporting against the UK Corporate Governance Code declare whether their material controls, reporting controls included, operated effectively (UK Corporate Governance Code, Provision 29). In the euro area, the ECB's supervisory priorities for 2026 to 2028 again name risk data aggregation and risk reporting as a material shortcoming.

The direction of travel is towards more granular data, more frequent submissions and more questions about individual outcomes. A supervisor who picks one figure from a return expects the firm to show where it came from and why it is right, quickly, and from evidence the firm already holds.

Why regulatory reporting errors pass every check

Most reporting controls test whether numbers agree: reconciliations to the ledger and to source systems, variance analysis between periods, and maker-checker review. Those controls catch errors of capture and errors of movement. They do not catch errors of treatment.

An error of treatment is a rule misread, a balance counted twice, or logic not updated after a rule change. It agrees with its sources every period, because it is applied consistently. It moves plausibly, because the business moves plausibly. So it passes every check and can persist for years.

Supervisors have said so directly. The PRA has twice warned firms not to take undue comfort from business-as-usual checking, because errors such as incomplete data or a misread requirement can persist unnoticed when large variances do not occur (Final Notices against Citigroup Global Markets in 2019 and Metro Bank in 2021). In 2026 a £99.9 million overstatement of an insurer's Own Funds passed through a data hand-off between two teams and the firm's checks for five quarters before a committee question exposed it.

The same pattern appears inside institutions' own estates. In one internal proof built on a client project, a figure labelled Loss Estimate was found to return raw exposure, with no probability of default or loss given default applied. It agreed with its sources every period, and failed against its objective at once.

Why the regulatory reporting process is hard to evidence

The process as documented is spread across policies, process documentation, the data catalogue, lineage tools and the controls register. The process as operated is data flowing through extraction jobs, transformations in application code, spreadsheets, manual adjustments and the judgement of people who have run it for years. Nothing holds the two together.

So when a supervisor, an auditor or a change programme asks how an outcome is produced, the firm reconstructs the answer by hand: people trace the data, open the code, ask the spreadsheet owner and write it up. The reconstruction is accurate on the day. It decays with the next change and is paid for again at the next question, the next mandate and the next audit. Lineage, where it exists, usually stops where the code begins, which is where the transformations that produce the figure sit.

Proving an outcome, not reconciling it

An outcome is not right or wrong in itself, only right for a purpose. The same client can carry several correct exposure figures on the same day, because credit risk, regulatory reporting and finance are each answering a different question. Reconciling one to another does not prove either, because it assumes a single true figure waiting to be found. Reconciliation is the right test only within a single question.

A reported outcome is proved by validating it against the objective it serves: name the question it answers and the rule that question requires, show the process that produced it, and run real data through that process to see whether it produces the reported figure. That test catches errors of treatment, because it checks the rule, not the agreement.

How xflow supports the regulatory reporting process

xflow builds process context: one executable view of the regulatory reporting process in which the documented and operated versions are held together, from the regulatory objective through policy, process steps, code, data, spreadsheets and manual adjustments to the reported figure. This is the golden copy of the process: one place where both versions are held and can be tested.

The transformations held in code are joined into the lineage, so the lineage can be run, not only read. Real data is run through the process to check whether it produces the reported figure. Gaps such as truncated lineage, undocumented transformations and missing business terms are found, assessed, fixed and written back to the data catalogue. Because the process's rules are held once and referenced by design, validation and evidence, the evidence for the next cycle comes from the same foundation rather than another reconstruction.

At a global tier one bank, lineage for regulatory reporting stopped where the application code began. xflow connected to the code repositories and produced the first analysis in hours. Transformation gaps were closed and written back to the bank's data governance layer, giving end-to-end lineage from policy to process to code to data.

xflow works alongside the regulatory reporting platform that produces the submission. That platform files the return. xflow works upstream of it, showing whether the process behind the return is the one the policy describes.

Questions about the regulatory reporting process

What is the regulatory reporting process?

The end-to-end chain from source data to submitted return: extraction, mapping, transformation, adjustment, review, approval and submission, together with the controls and evidence around each step.

What are FINREP and COREP?

The EU's harmonised supervisory reporting frameworks for banks: FINREP for financial information and COREP for capital and other prudential requirements, defined in the EBA's reporting standards.

Why do regulatory reporting errors go undetected?

Because most controls test whether figures agree with their sources and move plausibly. Errors of treatment agree and move plausibly, so they pass.

What do supervisors expect when they question an outcome?

That the firm can show how it was produced, from source data through each transformation and adjustment, and why the rule applied was the right one, from evidence it already holds.

Is reconciliation enough to prove a regulatory outcome?

No. Reconciliation shows that two figures agree. It proves correctness only when both answer the same question under the same rule.

How do insurers' reporting processes differ?

Solvency II and IFRS 17 add actuarial models, assumptions and judgements, often in spreadsheets owned by different teams, so a single disclosed outcome can depend on more hand-offs than a bank's return. The evidence question is the same.

Which outcome do you need to prove now?

Start with one critical outcome. We’ll show you how it is produced today, where that differs from what is documented, and how to close the gaps.

Start with one critical outcome