Why Financial Audit Discrepancies Happen and What Resolution Actually Means
A financial audit discrepancy is any difference between reported figures and the underlying source data — an invoice that does not match a purchase order, a general ledger account that drifts from the subledger, a payroll run that misses a pay code, or a bank reconciliation that fails to balance to the statement. The Connecticut Port Authority state audit released in 2025, for example, exposed multiple unreconciled intercompany balances and unsupported journal entries running into the millions of dollars. At Standard Chartered, regulators in 1994 found "discrepancies in expense claims" submitted by senior staff in Malaysia and the Philippines, which the bank acknowledged in an 18 July 1994 statement. These cases illustrate that discrepancies are rarely isolated arithmetic errors; they are usually symptoms of missing controls, broken processes, or pressured staff cutting corners. Resolution is therefore not just "fix the number" but rebuild the control environment so the same gap cannot reappear the following quarter.
Also worth reading: How Does a Financial Discrepancy Investigation Uncover Fraud? · How Can Financial Discrepancy Auditing Strengthen Public Accountability? · How Should Financial Close Discrepancy Testing Be Performed in 2026?
The Five Core Discrepancy Resolution Strategies Used in Modern Finance Teams
Most finance teams converge on five overlapping strategies. First, structured account reconciliation, where every balance-sheet account is matched to an independent source (bank statement, supplier statement, payroll register, contract schedule) on a defined cadence — typically monthly for high-risk accounts and quarterly for stable ones. Second, invoice-to-PO three-way matching, which compares the purchase order, goods receipt note, and supplier invoice before any payment is released. Third, payroll audits, where a sample of pay runs is recomputed against time-and-attendance data to catch phantom employees, misclassified workers, and benefit overpayments. Fourth, continuous monitoring using AI agents that scan 100 percent of transactions rather than the 5–10 percent a human samples. According to PwC's 2025 finance operating model research, AI agents now allow controllers to review every journal entry, every invoice, and every reconciliation break in near real time, rather than relying on month-end sampling. Fifth, formal dispute escalation frameworks — the kind used by the UK's Financial Dispute Resolution Centre (FDRC), which mediates cases where documentation is contested by both parties.
How to Resolve a Discrepancy Step by Step
The practical workflow has six phases and should be written into the company's audit resolution policy. Phase one is identification: a reconciliation break, an exception report, or an auditor's finding flags the variance. Phase two is isolation, where the transaction is frozen, supporting documents are gathered, and the responsible controller is assigned. Phase three is root-cause analysis — was this a timing difference, a posting error, a fraud indicator, a system mapping failure, or a deliberate misstatement? Phase four is correction, with an adjusting journal entry that is approved by someone independent of the preparer. Phase five is documentation: a reconciliation workpaper showing the cause, the correction, the approver, and the date, retained for at least seven years to satisfy most regulators. Phase six is preventive control, where the underlying process is redesigned — a new validation rule, an approval threshold change, or a system automation. Skipping phase six is the single most common reason the same discrepancy reappears at the next audit.
Comparing Manual, Automated, and AI-Driven Resolution Approaches
The table below compares the three dominant approaches finance teams use in 2026. Numbers reflect typical mid-market deployments, not theoretical benchmarks.
| Feature | Manual Reconciliation | Automated Reconciliation Software | AI-Driven Reconciliation Agents |
|---|---|---|---|
| Transaction coverage | 5–10 percent sample | 80–95 percent of high-volume accounts | 100 percent of in-scope transactions |
| Time to close a typical month | 8–12 business days | 4–6 business days | 1–3 business days |
| Cost per 1,000 reconciliations | $1,200–$2,000 | $250–$500 | $90–$180 |
| Detection of unusual entries | Low (depends on human focus) | Medium (rule-based flags) | High (anomaly detection across patterns) |
| Typical error rate after 12 months | 2–4 percent of accounts | 0.5–1.5 percent | Under 0.3 percent |
| Up-front implementation time | None | 6–12 weeks | 10–20 weeks |
| Best fit | Very small entities, low transaction volume | Mid-market firms with ERP in place | Enterprises and audit-heavy regulated firms |
Common Mistakes That Make Discrepancies Worse
Several recurring mistakes turn a 30-minute fix into a 30-day investigation. Treating the symptom rather than the cause is the most expensive: a controller posts a correcting journal entry, closes the reconciliation, and never asks why the variance happened in the first place. Another is letting unsupported "plug" entries accumulate. A plug is a balancing figure posted to make an account agree with no documented explanation, and auditors specifically flag plugs because they are statistically the strongest predictor of material misstatement. The Department of Defense struggled with this across FYs 1998–2010, when its financial statements were either unauditable or carried no audit opinion because reconciliations could not be supported. Allowing the same person to prepare and approve a correction is a third mistake — it removes the segregation of duties that prevents both error and fraud. Finally, delaying escalation beyond 30 days is dangerous: the older a discrepancy, the harder it is to investigate because source documents may be archived, staff may have left, and the trail of compensating entries becomes tangled.
When to Escalate a Discrepancy Beyond Routine Resolution
Not every variance belongs in the same workflow. As a practical rule of thumb, any single discrepancy above 0.5 percent of revenue, any suspected regulatory breach, and any item with fraud indicators (round-number entries, weekend postings, after-hours system access, missing documentation) should be escalated immediately to the head of internal audit and, where material, to the audit committee. The Nigerian Senate's 2024 order to the Nigeria Customs Service over an N62.2 billion query shows how discrepancies at scale quickly move out of the finance department's hands and into legislative and public scrutiny. Internationally, the 2008 Global Financial Crisis produced several cases where small reconciliation failures in mortgage servicing files turned into systemic losses once aggregated across portfolios — a reminder that the size of a single discrepancy matters less than the pattern across many of them. Escalation thresholds should be written into the company's fraud response plan, not left to individual judgment.
Building a Sustainable Resolution Framework for 2026 and Beyond
The most resilient framework in 2026 combines three layers. The first layer is preventive: system-enforced controls such as three-way match in the ERP, mandatory field validation, and dual approval on journal entries above a set dollar threshold. The second layer is detective: continuous reconciliation and AI-driven anomaly detection that runs daily, not monthly. The third layer is responsive: a documented, time-bound resolution process with named owners, escalation triggers, and a quarterly review of repeat discrepancies by the audit committee. Organizations that operate all three layers typically report a 60–80 percent reduction in audit adjustments year-on-year, while those that rely on a single layer — usually detective monitoring alone — see diminishing returns as exceptions accumulate faster than they can be investigated. The cost of building the three-layer framework ranges from roughly $50,000 for a small cloud-based deployment to $2 million-plus for a regulated enterprise, but the payback period is generally under 18 months once the reduction in audit fees, restatement risk, and staff overtime is counted.
The Honest Limits of Discrepancy Resolution
It is worth being direct about what good resolution cannot do. It cannot rescue a company that has been systematically falsifying its books — at that point the work belongs to forensic accountants, regulators, and sometimes law enforcement. It cannot eliminate every timing gap in a complex multi-entity group where intercompany balances routinely disagree by thousands of dollars on the last day of the month, though it can shrink those gaps by an order of magnitude. And it cannot replace professional skepticism: a clean reconciliation produced too quickly, or signed off without challenge, is itself a red flag. The Connecticut Port Authority audit, the Standard Chartered expense-claims case, and the CapRadio board resignations of 2024 all share a common thread — the controls existed on paper, but the people applying them did not push back when the numbers did not make sense. The most effective discrepancy resolution strategy is therefore the one that gives trained professionals the time, the tools, and the authority to question what looks finished.