Post Office Horizon Scandal: Evidence and Ethics for Accountants

Public record checked: 20 October 2026. This article distinguishes court decisions, legislation and the Post Office Horizon IT Inquiry. Practical examples below are fictional training scenarios, not further findings about any person or organisation.

A balance on a screen is a starting point for investigation. It does not, by itself, establish that cash has disappeared or that the person responsible for the account has acted dishonestly. The Post Office Horizon scandal demonstrates the human consequences of losing that distinction.

For accountants, the enduring questions concern evidence: how a number was produced, whether it can be reconciled to independent records, what might contradict it, and whether people can challenge it without being dismissed. These questions apply to an established accounting system just as much as a newly introduced automated tool.

What happened in the Post Office Horizon scandal?

Horizon was a computerised accounting system used in Post Office branches. Disputes about apparent branch shortfalls led to civil litigation and appeals against criminal convictions. These were distinct proceedings with different questions to decide.

In the December 2019 Bates Horizon Issues judgment, the High Court examined the system's reliability, including bugs, errors, defects and the capacity for remote intervention. Its findings undermined the assumption that a recorded shortfall necessarily demonstrated a real loss at a branch.

The Court of Appeal subsequently considered 42 appellants in Hamilton and others v Post Office Limited, decided on 23 April 2021. It quashed the convictions of 39 appellants. The other three appeals were dismissed because the reliability of Horizon data was not essential to those prosecution cases. That distinction matters: an account of the judgment should describe the cases the court actually decided.

The court's published summary explains that, in the relevant Horizon cases, apparent shortages could arise from system problems rather than missing cash or stock. It also describes serious failures of investigation and disclosure. These were judicial conclusions about the evidence and prosecutions before the court, not simply a later disagreement about software quality.

What has been resolved, and what remains separate?

The Post Office (Horizon System) Offences Act 2024 provided for qualifying convictions in England, Wales and Northern Ireland to be quashed. Its statutory conditions matter; it should not be described as overturning every conviction involving every Post Office system. Scotland legislated separately. Government management information also distinguishes convictions within scope from cases outside it.

The Inquiry's first final-report volume, published on 8 July 2025, concerns human impact and redress. In his 8 July 2026 update, Sir Wyn Williams said five further volumes were being prepared and explained the process for giving those facing proposed criticism an opportunity to respond. The publications page checked for this article did not list those remaining volumes as published.

Quashing a conviction, awarding redress, publishing an inquiry finding and determining criminal responsibility are different processes. An educational article should not use one as a substitute for another.

The organisations involved have also made public acknowledgements. In an October 2025 joint statement, the Department for Business and Trade, Fujitsu Services Limited and Post Office Limited apologised for the damage their organisations caused and discussed restorative justice. That acknowledgement belongs alongside the public findings; it does not replace the detail of the judgments or the Inquiry's work.

Why system output is not independent corroboration

Suppose a ledger shows a £4,000 shortage and a management dashboard repeats that number. If the dashboard is fed by the same ledger, the two outputs are one evidential chain. Agreement between them does not establish that the underlying transactions were complete and accurate.

A reviewer needs to ask where the data originated, which interfaces transformed it, who could amend it, and whether corrections are visible. External settlement records, independently obtained bank information and physical records may answer different parts of the question. Their relevance depends on the transaction being tested.

This is a general control lesson rather than a claim that any particular reconciliation would have resolved every Horizon case. A reliable review considers both the accounting assertion and the limitations of the evidence available to support it.

The same care applies to exception reports. A report can be accurately generated from an incomplete population. Before relying on a clean exception list, establish what it includes, what it excludes and whether missing or rejected records are visible elsewhere.

How should an accountant investigate an unexplained discrepancy?

A useful investigation separates the observation from the explanation. “The system reports a shortfall” records an observation. “Someone removed the money” proposes a cause. Moving between those statements requires evidence.

1. Define the difference. Record the amount, account, period and reconciliation boundary. Check whether the issue concerns cash, stock, settlement timing or an accounting entry.
2. Preserve the original records. Retain relevant exports, logs, correspondence and versions before adjustments overwrite the position being investigated. Use appropriate access and retention controls.
3. Test competing explanations. Consider timing, duplicate processing, failed interfaces, incorrect opening balances, authorised amendments and misconduct. A plausible first explanation should not close the enquiry.
4. Seek independent evidence. Trace selected items to records outside the disputed data flow where possible. Document how each item supports or contradicts the proposed explanation.
5. Record uncertainty and ownership. Identify unresolved points, the person responsible for the next check and the date for review. An unresolved difference is not evidence that a preferred explanation is correct.

Materiality should guide the financial investigation, but potential harm to a person also deserves attention. A relatively small reported shortage can have serious employment, contractual or personal consequences if it is treated as proof of wrongdoing.

Worked example: the difference that survives a reconciliation

Fictional training scenario. A branch account shows a £4,000 shortage after a software update. The cash count agrees with the bank deposit slip. A supervisor suggests charging the amount to the branch manager because the same shortage appears in the monthly dashboard.

The finance reviewer maps both reports and finds they use the same transaction table. She preserves the original export, obtains the bank settlement record independently and compares transaction identifiers across the interface. A settlement batch appears twice in one feed but once in the bank record.

The appropriate response is to investigate that mismatch, establish its scope and obtain approval for any correcting entry. It would be premature to accuse the manager, but equally premature to conclude that every difference in the organisation has the same technical cause.

The review file should show the evidence obtained, the hypotheses rejected, the correction approved and the checks performed on other affected accounts. Closing the numerical difference is only part of closing the control incident.

Ethical escalation requires a usable route

A policy encouraging staff to speak up has limited value if the person receiving a concern is also the person defending the original conclusion. Teams need an escalation route capable of obtaining technical help and authorising independent review.

For example, a concern log can distinguish the observed discrepancy, the evidence supporting the current explanation, contradictory information and the decision required. That structure makes disagreement reviewable. It also prevents a general assurance such as “the system has been checked” from displacing a specific unresolved issue.

IFAC's discussion of professional scepticism and an inquiring mind distinguishes the audit concept from the questioning approach expected more broadly of professional accountants. The practical habit is to assess information critically and remain willing to revise a conclusion when contrary evidence emerges.

Staff should know where to escalate if their concern remains unresolved and how to preserve confidentiality. The appropriate route depends on their role, professional obligations and the organisation's procedures; a general case study cannot determine every individual's reporting duties.

Questions for the next system-generated exception

• Can we trace this number to the underlying transaction and explain every transformation?
• Are the apparently supporting reports independent, or do they repeat the same source?
• Have we checked access, amendments, known incidents and rejected transactions?
• What evidence would disprove our current explanation?
• Have we heard and documented the affected person's account?
• Can someone independent review an unresolved disagreement?

The most useful lesson from Horizon is an evidential discipline: establish what a number proves before deciding what action it justifies. That discipline protects the quality of financial reporting and the people affected by financial decisions.

For related learning, read our guides to fraud risk and internal controls and Carillion's accounting and audit lessons. Our Manchester City financial-rules explainer applies the same care in distinguishing different proceedings and findings.