External auditors and regulatory examiners are not there to catch mistakes before they become problems. They are there to verify that the numbers you reported are correct. When they find a discrepancy, the investigation that follows is expensive, time-consuming, and visible in ways that internal detection is not. The fact that most discrepancies are detectable internally, before the external reviewer arrives, is one of the less-discussed opportunities in audit preparation.
The challenge is detection timing. Most audit preparation processes are front-loaded toward assembly and back-loaded toward substantiation. The team spends time gathering documents and organizing working papers, then the external reviewer arrives and discovers discrepancies that have been sitting in the working papers for weeks. The discrepancy was present during the entire preparation phase. Nobody caught it because nobody was systematically checking the connection between figures and their sources in real time.
What source discrepancies actually look like
A source discrepancy is any case where a figure in a working paper does not match the source document it is supposed to trace to. These fall into several distinct categories, each with different causes and different remediation paths.
Numerical mismatches are the most visible type: the working paper says $3.7M and the board package says $3.6M. Sometimes this reflects a legitimate adjustment that was not documented; sometimes it reflects a transcription error; sometimes it reflects a version confusion where the working paper was updated to the final figure but the source citation still points to an earlier version.
Unit and rounding discrepancies are common and frequently misdiagnosed. A figure expressed in thousands in the working paper and in full dollars in the source document is the same number, but without automated matching that accounts for unit conversions, it looks like a discrepancy. Conversely, a genuine rounding discrepancy where someone rounded a subtotal before including it in a larger calculation can look like a unit error. Distinguishing between these requires context that manual review often does not capture consistently.
Labeling mismatches occur when the description of a figure in the working paper does not match its description in the source document. A line labeled "Q3 Net Revenue" in the working paper traces to a line labeled "Third Quarter Total Revenue, net of returns" in the board package. The figures match, but the labels differ. Under careful external review, this type of mismatch can prompt questions about whether the figures are truly equivalent or whether there are definitional differences.
Citation drift is perhaps the most insidious type. A working paper was built using version three of a board package. The final approved version is version five. The figures in the working paper were reconciled to version five before close, but the citations were not updated. The working paper cites version three for figures that actually trace to version five. No numerical discrepancy exists, but the citation is wrong, and an external reviewer who checks the cited document against the working paper figure will find a discrepancy.
Why discrepancies persist through internal review
Internal review processes are generally good at catching numerical errors that are apparent from the working papers themselves: mathematical errors in calculations, transposition errors in significant figures, obvious mismatches between totals and their components. They are less good at catching discrepancies that require checking the working paper against its source documents.
Source-checking in manual processes is sampling-based. A reviewer checks a subset of figures against their sources, focusing on the most material items and the areas of highest risk. This is reasonable given time constraints. It means that a discrepancy affecting a figure that did not fall in the review sample may not be caught until the external auditor independently selects that figure for testing.
The other factor is that internal reviewers often have contextual knowledge that external reviewers lack. When an internal reviewer encounters a figure that does not quite match its source, they may know from memory why the difference exists and resolve it mentally without flagging it. An external reviewer has no such context and must follow the documented chain wherever it leads, including into discrepancies that internal teams considered resolved but did not document.
The detection timing advantage
The most significant cost of discovering a discrepancy during external review, rather than during internal preparation, is not the remediation cost. It is the investigation cost. An external auditor who finds a discrepancy must determine whether it represents a transcription error, a judgment call, a deliberate adjustment, or something more significant. That determination requires time from both the internal team and the external reviewer, and it extends fieldwork.
A discrepancy detected internally during the working paper update process is a different category of finding. The person who updated the working paper cell knows immediately whether the figure changed because a source document was revised, because an adjustment was applied, or because of a data entry error. Resolution is fast, context is available, and the correction can be made before the citation chain is even finalized.
This is the core value of real-time source tracing during working paper preparation. When the link between a working paper cell and its source is checked automatically at the moment of cell update, discrepancies surface at the point when they are cheapest and easiest to resolve. The same discrepancy detected six weeks later, during fieldwork, costs significantly more to investigate and carries meaningfully more risk.
What automated source tracing actually checks
Automated source tracing for discrepancy detection is not the same as automated working paper review. The system does not evaluate whether figures are materially correct or whether accounting treatments are appropriate. Those remain professional judgments.
What the system checks is a specific and bounded question: does this working paper figure match the value in its cited source document, accounting for unit conventions, rounding conventions, and label variations? If yes, the citation is confirmed. If no, the discrepancy is flagged with the specific nature of the mismatch (numerical, unit, label, citation drift) and the confidence level of the check.
Flagged discrepancies go to the working paper preparer for resolution before the document is finalized. This shifts the burden of discrepancy detection from external reviewers who have limited context to the internal team who built the working paper and know the document set. It also creates an audit log of every discrepancy that was detected and resolved, which is itself useful documentation for explaining the working paper to an external reviewer.
The pre-fieldwork versus in-fieldwork cost differential
Consider the difference in resolution cost between these two scenarios: a discrepancy identified by a Pramaana flag during working paper preparation, versus the same discrepancy identified by an external auditor during fieldwork.
In the first scenario, the working paper preparer sees the flag, checks the source document, confirms the discrepancy is a citation error (the document reference was not updated when the figure was reconciled to the final board package), updates the citation, and the discrepancy is resolved in under fifteen minutes.
In the second scenario, the external auditor identifies the discrepancy during fieldwork, raises a request for explanation to the internal team, the internal team locates the person who built the working paper, reconstructs the context of why the discrepancy exists, prepares a written response, the external auditor evaluates the response, determines it is satisfactory, and closes the item. The resolution takes two to three days and involves senior time from both teams.
The same discrepancy, discovered at different points in the cycle, has fundamentally different cost profiles. Systematic pre-fieldwork detection is not about preventing discrepancies from existing. Working papers will always have discrepancies that require resolution. It is about shifting detection timing to where resolution is cheap, rather than where it is expensive.
For internal audit teams who want to reduce the number of external auditor findings that originate from evidence chain issues, real-time source tracing during working paper preparation is the most direct intervention available. The Pramaana trace engine is built specifically to provide this detection capability alongside its provenance chain construction, so that discrepancy flags and citation links are maintained in the same system and the same workflow.