Records, Evidence, and Tax Data Quality

Introduction and Context

Tax compliance is not established merely because a return has been filed or a tax payment has been made. It is established when the taxpayer, adviser, or organisation can demonstrate how reported figures were derived, what evidence supports them, who prepared and reviewed the work, and whether the records remain reliable and accessible. Records therefore perform two functions: they support the preparation of accurate tax returns, and they provide an audit trail through which decisions can be independently examined.

This chapter examines the practical foundations of tax records and evidence. It focuses on substantiation, reconciliation, documentation, data quality, confidentiality, electronic records, and retention. Although exact legal requirements differ by jurisdiction and tax type, the underlying professional principles are widely applicable. Tax professionals must always confirm the rules applicable to the taxpayer, period, transaction, and jurisdiction rather than relying on general practice.

A useful distinction is between a transaction and its evidence. A transaction is an economic event, such as a purchase, sale, payment, loan, or payroll entry. Evidence is the information that demonstrates that the event occurred, identifies its parties and terms, establishes its amount and timing, and explains its tax treatment. A ledger entry without supporting evidence may be insufficient; an invoice without an accounting entry may also be incomplete. Reliable compliance connects the two.

Learning Objectives

By the end of this chapter, you should be able to:

1. Explain the purposes and characteristics of tax records and audit trails.
2. Distinguish source documents, accounting records, working papers, and tax filings.
3. Apply a substantiation framework to common tax claims.
4. Reconcile tax data to accounting records, bank evidence, and returns.
5. Identify and correct completeness, accuracy, validity, timeliness, and consistency problems.
6. Design proportionate paper and electronic record systems.
7. Protect confidential tax information and manage access responsibly.
8. Apply defensible retention, version-control, review, and disposition practices.

1. The Architecture of Tax Evidence

1.1 Four layers of the record system

A dependable record system normally contains four connected layers:

The layers should be traceable in both directions. A reviewer should be able to start with a tax-return figure and trace it to a working paper, ledger account, and source documents. Conversely, a source document should be traceable to the accounting and tax treatment applied to it. This is often called a complete audit trail or chain of custody for information.

1.2 Source documents and evidential value

A source document is most useful when it answers five questions:

• Who were the parties?
• What was supplied or received?
• When did the event occur?
• What amount and terms applied?
• Why is the tax treatment appropriate?

A document’s existence does not automatically make it persuasive. An undated spreadsheet created after an enquiry may have less evidential value than a contemporaneous invoice, contract, delivery record, and bank payment considered together. Evidence is stronger when it is contemporaneous, independent, internally consistent, attributable to an identifiable person or system, and capable of being verified.

Common source documents include sales invoices, purchase invoices, receipts, contracts, purchase orders, delivery notes, payroll records, travel logs, inventory records, loan agreements, property documents, bank statements, payment processor reports, customs records, and correspondence with tax authorities. Digital records may include structured transaction files, system logs, electronic approvals, metadata, and archived email.

1.3 Substantiation is a reasoning process

Substantiation is not simply collecting a large quantity of documents. It is the process of demonstrating that a tax treatment follows from relevant facts and applicable rules. For example, a business-expense claim may require evidence of payment, business purpose, amount, date, and—where mixed use exists—a reasonable allocation between business and private use.

A professional substantiation file should normally contain:

1. the underlying source document;
2. a statement of the relevant facts;
3. the tax rule or policy applied;
4. the calculation or allocation, if required;
5. evidence of review and approval; and
6. explanations of unusual items, estimates, or uncertainties.

The standard is not that every record must be perfect. Rather, the records should be sufficient, relevant, reliable, and proportionate to the risk and value of the matter.

2. Audit Trails and Documentation Discipline

An audit trail records the movement of information from origin to conclusion. At a minimum, it should show the source, transformation, decision, reviewer, date, and final use of material information. For a tax adjustment, this might mean linking a trial-balance account to a reconciliation, identifying excluded or added items, recording the legal or policy rationale, and linking the final amount to the return.

A sound audit trail has the following characteristics:

• Traceability: each material figure can be followed to its origin and destination.
• Chronology: preparation, review, amendment, and submission events are dated.
• Attribution: it is clear who created, changed, reviewed, or approved a record.
• Integrity: records are protected from unauthorised alteration or deletion.
• Clarity: another competent person can understand the work without relying on undocumented memory.
• Reproducibility: the result can be recreated using the retained inputs, formulas, and assumptions.

Documentation should distinguish facts from judgements. “Invoice dated 12 March records payment of 1,200” is a fact supported by a document. “The cost is wholly deductible” is a conclusion requiring a rule, facts, and professional reasoning. Where a position is uncertain, the file should record the uncertainty, alternatives considered, materiality, and approval decision.

A practical file index might include: client and period identifiers; source-document register; bank and ledger reconciliations; tax adjustments; estimates; correspondence; review notes; filing copy; payment evidence; and post-filing amendments. Consistent naming conventions—such as client, tax type, period, document type, and version—reduce retrieval errors.

Figure: A bidirectional audit trail linking transactions, evidence, accounting records, and tax reporting.

3. Reconciliation and Control of Tax Data

3.1 The purpose of reconciliation

Reconciliation compares two independent or differently generated records and explains any difference. It is a control, not merely a clerical exercise. Typical reconciliations include:

• bank statements to cash ledgers;
• sales invoices to revenue ledgers;
• purchase records to input-tax claims;
• payroll reports to payroll-tax filings;
• fixed-asset registers to depreciation claims;
• inventory records to cost-of-sales amounts; and
• accounting profit to taxable profit.

A reconciliation should identify the populations compared, the period, the responsible preparer, the date performed, the differences found, and the resolution or outstanding status of each difference. “Agrees” without showing the basis of agreement is weak documentation.

3.2 A basic reconciliation model

For a balance that changes during a period, the general roll-forward relationship is:

“Opening balance” is the verified balance at the start of the period; “additions” are increases recorded during the period; “reductions” are decreases; and “other adjustments” include reclassifications, foreign-exchange movements, corrections, or approved estimates. The identity is useful only when the components are defined consistently and relate to the same entity, period, and measurement basis.

For example, if a tax-payable account begins at 4,000, contains current-period charges of 18,500, payments of 16,000, and a credit adjustment of 500, the closing balance is:

The result must then be compared with the authority’s account or payment records. A mathematical agreement does not prove that the original charge or payment was correctly classified; it only confirms the roll-forward arithmetic.

3.3 Exception-based review

Efficient review concentrates on exceptions rather than treating every item identically. Useful exception indicators include duplicate invoice numbers, missing tax identifiers, unusual tax rates, transactions posted outside the period, round-number journal entries, negative quantities, dormant suppliers becoming active, and manual overrides. Exception rules should be documented, tested, and periodically updated so that efficiency does not become a substitute for judgement.

4. Tax Data Quality

Tax data quality is multidimensional. A dataset may be accurate in individual fields yet incomplete overall, or complete but poorly classified. The principal dimensions are:

These dimensions should be converted into controls. For example, completeness may be tested by comparing sequential invoice numbers, payment-provider totals, and bank deposits. Validity may be tested through required fields, date ranges, tax-code lists, and supplier identifiers. Accuracy may require sample inspection against original documents. Consistency may require automated comparison between the general ledger and tax return working papers.

4.1 Data-quality correction

Corrections should preserve the original information and explain the change. A robust correction record identifies the original value, corrected value, reason, author, date, reviewer, and affected reports. Deleting an erroneous record without leaving a trace weakens integrity and may create the appearance of concealment. In accounting systems, a reversing or adjusting entry with an explanatory reference is generally more defensible than an invisible overwrite.

4.2 Estimates and allocations

Tax work often involves estimates or allocations—for example, business use of a vehicle, mixed-use premises, provisions, or depreciation periods. An allocation should identify:

• the total cost or population;
• the allocation basis;
• the period measured;
• the evidence supporting the basis;
• the calculation; and
• the review date.

A precise-looking percentage is not necessarily reliable. A reasonable, documented method applied consistently may be stronger than an unsupported figure with several decimal places. If circumstances change, the allocation should be reassessed rather than carried forward automatically.

5. Paper and Electronic Records

5.1 Paper records

Paper records should be legible, organised, protected from loss or deterioration, and indexed so that they can be retrieved. Originals may be important where law, contractual terms, or evidential circumstances require them. Scanning is useful, but a scan should be checked for completeness, readability, page order, and correct association with the transaction.

5.2 Electronic records

Electronic records require controls over access, alteration, backup, migration, and retrieval. Important controls include:

• role-based access and least privilege;
• multi-factor authentication for sensitive systems;
• immutable or protected audit logs where appropriate;
• version control and controlled templates;
• regular backups, including an off-site or otherwise independent copy;
• tested restoration procedures;
• malware and endpoint protection;
• documented export formats and system dependencies; and
• periodic review of inactive accounts and external access.

A PDF may preserve visual appearance but not necessarily the underlying data, formulas, metadata, or audit history. For important spreadsheets, retain both the locked reporting copy and the underlying source data, formulas, assumptions, and version history where practicable. Macros and automated transformations should be documented and independently checked when material.

Cloud storage does not eliminate recordkeeping responsibility. The organisation remains responsible for access permissions, supplier due diligence, location and transfer issues, service continuity, retention settings, and the ability to retrieve records in a usable format.

5.3 Technology and AI

Optical character recognition, automated data extraction, duplicate detection, and rules-based validation can improve speed and consistency. AI tools may assist with classification, summarisation, anomaly identification, or document search. However, generated outputs are not evidence merely because a system produced them. Human review remains necessary for material tax judgements, unusual transactions, legal interpretation, privacy, and completeness.

AI use should be governed by approved tools, secure input rules, access controls, logging, testing, and clear accountability. Confidential taxpayer information should not be entered into an unapproved public system. Professionals should retain the original documents and record material prompts, transformations, validation steps, and reviewer conclusions when AI materially affects the work.

6. Confidentiality, Privacy, and Ethical Stewardship

Tax records commonly contain identity information, bank details, payroll data, business strategies, and commercially sensitive contracts. Confidentiality is both an ethical duty and, in many jurisdictions, a legal or regulatory obligation. Access should be based on a legitimate business need, not convenience or status.

Core safeguards include data minimisation, secure transmission, encryption where appropriate, clean-desk practices, controlled printing, secure disposal, incident reporting, and documented authorisation for disclosure. Emailing a spreadsheet containing unnecessary personal data is avoidable risk. Redaction should remove the underlying data rather than merely covering it visually.

Ethical recordkeeping also requires resistance to manipulation. A professional must not backdate a document, fabricate a business purpose, destroy an adverse record, conceal an error, or create a misleading audit trail. If a material error is discovered, it should be assessed, documented, corrected through an authorised process, and communicated to the appropriate person or authority where required. Confidentiality does not justify concealing unlawful conduct or preventing a legally required disclosure.

7. Retention, Retrieval, and Disposition

Retention is governed by the interaction of tax law, company law, employment and privacy requirements, contractual duties, litigation holds, and professional policies. There is no universally correct retention period. An organisation should identify the applicable rule for each record category and retain records for at least the required period, subject to longer periods where a dispute, investigation, audit, amendment period, or legal hold exists.

A retention schedule should specify the record category, responsible owner, retention trigger, minimum period, storage location, access rules, and authorised disposal method. The trigger may be the end of a tax period, filing date, transaction completion, employment termination, or closure of an enquiry. Records should not be destroyed merely because they are inconvenient or because a system is being replaced.

At the end of the approved period, disposition should be authorised and documented. Secure destruction may involve cross-cut shredding, certified disposal, cryptographic erasure, or verified deletion from systems and backups, subject to technical realities. Where deletion from backups is impracticable, access and restoration controls should prevent routine use, and the policy should explain the treatment.

8. Applied Case: The Missing Input-Tax Evidence File

This is an illustrative case.

Background

Harbour Workshop Ltd, a small repair business, claimed input tax on purchases for a quarterly return. Its bookkeeping system showed purchase invoices totalling 48,000, but the electronic folder contained only 39,500 of invoices. The accounts payable assistant explained that some documents were in email, some were paper receipts, and some had been uploaded twice.

Problem

The return had been prepared from the ledger without a documented reconciliation to supplier statements, bank payments, or the tax-code report. Several expenses related partly to the owner’s private vehicle. The business also used a shared cloud folder with no restricted permissions or version history.

Analysis

The missing documents created a completeness and substantiation risk. The duplicate uploads created a potential accuracy risk. The mixed-use vehicle costs required an allocation rather than automatic full recovery. The absence of a review trail meant that the preparer could not demonstrate how the reported amount had been assembled.

Approach and solution

The adviser first froze the relevant folder and exported the ledger, tax-code report, supplier statements, and bank transactions. Each purchase was assigned a unique reference. The team matched invoices to ledger entries and payments, removed duplicates through authorised adjusting entries, and placed unsupported items in an exception list. Suppliers were asked for copies of missing invoices; paper receipts were scanned and checked for legibility. Vehicle costs were separated and allocated using a documented business-use log. The adviser then prepared a reconciliation showing the original claim, supported items, excluded items, and unresolved items. A manager reviewed the file before any amendment or final submission.

Access to the cloud folder was restricted by role, multifactor authentication was enabled, and a naming convention and retention schedule were introduced. The business also adopted a monthly close checklist requiring invoice completeness testing and review of unusual tax codes.

Outcome

The corrected claim was lower than the original ledger-based amount, but it was defensible and reproducible. The business avoided relying on unsupported records and improved its process for future periods. The case illustrates that a smaller, well-supported claim is professionally stronger than a larger claim that cannot be substantiated.

Learning points

1. A ledger total is not a substitute for source evidence.
2. Reconciliation should be performed before, not after, a compliance problem arises.
3. Missing, duplicate, and mixed-use items require different corrective responses.
4. Access control and version history are part of evidential integrity.
5. Review documentation should explain both the conclusion and the exceptions.

9. Best-Practice Control Framework

A proportionate control framework can be organised around the record lifecycle:

1. Capture: obtain complete source documents at or near the transaction date.
2. Classify: apply the correct account, tax code, period, entity, and cost centre.
3. Validate: perform format, duplicate, reasonableness, and authority checks.
4. Reconcile: compare independent records and investigate exceptions.
5. Review: document judgements, estimates, corrections, and approval.
6. Submit: retain the final return, computation, payment evidence, and acknowledgement.
7. Protect: control access, preserve integrity, and back up records.
8. Retain and dispose: follow the applicable schedule and legal holds.

The framework should be risk-based. High-value, unusual, related-party, cross-border, cash, and manually adjusted transactions normally merit deeper review than routine low-value items. Nevertheless, routine processes should not be ignored: repeated small errors can become material in aggregate.

Challenges and Limitations

Record systems face practical limitations. Small organisations may lack dedicated compliance staff, and legacy systems may not export complete audit logs. Digital preservation can be complicated by obsolete software, encrypted files, supplier portals, and changing formats. Excessive documentation can create cost and information overload, while insufficient documentation creates audit and ethical risk.

Controls can also produce false confidence. A system may show that a field is completed without proving that the field is truthful. Automated matching can miss deliberate collusion or unusual but legitimate transactions. Reconciliations may agree because the same error was entered in two systems. For these reasons, data controls should combine automated tests, independent evidence, professional scepticism, and periodic process review.

Professional Implications

Tax professionals are not only preparers of calculations; they are custodians of evidence. Their work should be understandable to a successor, defensible to an auditor, and respectful of the taxpayer’s rights. Good records reduce disputes, support timely responses, improve cash and filing management, and help identify tax risks before they become penalties or reputational damage. They also support continuity when staff leave or systems change.

Chapter Summary

Tax compliance depends on a connected evidence system linking transactions, source documents, accounting records, tax working papers, and filings. A complete audit trail provides traceability, chronology, attribution, integrity, clarity, and reproducibility. Substantiation requires evidence of the event, amount, timing, purpose, parties, and tax treatment, including documented allocations and estimates.

Reconciliation is a control that compares independent records and explains differences. Tax data should be assessed for accuracy, completeness, validity, consistency, timeliness, uniqueness, integrity, and relevance. Electronic and paper records require different but complementary safeguards, including access control, version management, backups, retrieval testing, and secure disposal. Confidentiality and ethical stewardship prohibit manipulation, unauthorised disclosure, and destruction of adverse evidence. Retention must follow applicable legal requirements, disputes, and documented schedules. Technology and AI can improve processing, but their outputs require human oversight and do not replace original evidence or professional judgement.

Key Terms

Audit trail: A traceable record of the source, processing, review, and reporting of information.

Source document: Original or primary evidence of a transaction or event.

Substantiation: Demonstration that a tax claim or treatment is supported by relevant facts and evidence.

Working paper: A document explaining a calculation, adjustment, judgement, or review.

Reconciliation: Comparison of records to identify and explain differences.

Data quality: The fitness of data for its intended compliance or reporting purpose.

Completeness: The extent to which all relevant records and fields are present.

Validity: Conformity with required formats, rules, and permitted values.

Data integrity: Protection of information from unauthorised or unexplained alteration.

Traceability: Ability to follow information to and from its source.

Version control: A method of identifying and preserving successive changes to a record.

Retention schedule: An authorised plan stating how long records are kept and when they may be disposed of.

Legal hold: A requirement to preserve records because of anticipated or actual litigation, investigation, or enquiry.

Access control: Rules determining who may view, change, or disclose information.

Least privilege: Granting users only the access necessary for their role.

Allocation: Assignment of a shared cost or amount between relevant categories using a documented basis.

Exception: A transaction or data item requiring investigation because it falls outside an expected rule or pattern.

Confidentiality: The duty to prevent unauthorised access or disclosure of information.

Review Questions

5 Short Answer Questions

1. What are the four main layers of a tax evidence system?
2. State four characteristics of a strong audit trail.
3. Why is reconciliation more than a mathematical comparison?
4. Name five dimensions of tax data quality.
5. Why should corrections preserve the original record and document the change?

5 Long Answer Questions

1. Explain how source documents, accounting records, working papers, and tax filings should be linked in a complete audit trail.
2. Discuss the evidential requirements for a mixed-use business expense and explain why an allocation may be necessary.
3. Compare the principal controls needed for paper records and electronic records.
4. Explain how confidentiality, access control, and ethical duties affect the management of tax records.
5. Describe how a retention schedule should operate when an audit, dispute, or legal hold is present.

3 Analytical / Application Questions

1. A company’s sales ledger agrees with its tax return, but bank deposits are 8% lower and several invoice numbers are missing. Design a sequence of investigations and controls to determine whether the return is complete and accurate.
2. A tax adviser uses an AI tool to extract invoice data and classify tax codes. Evaluate the risks and propose a governance process that preserves confidentiality, accuracy, accountability, and auditability.
3. A taxpayer has complete invoices but no evidence of business purpose for a set of travel claims. Analyse the substantiation weakness and recommend evidence and review procedures for the current and future periods.

Multiple Choice Questions with Answers

Question 1

Which record provides the primary evidence that a transaction occurred?

A. A tax return summary

B. A source document such as an invoice or receipt

C. A general compliance policy

D. A staff training schedule

Correct Answer: B. A source document such as an invoice or receipt

Question 2

What is the principal purpose of a reconciliation?

A. To delete unusual transactions

B. To compare records and explain differences

C. To replace source documents

D. To guarantee that every tax position is legally correct

Correct Answer: B. To compare records and explain differences

Question 3

Which data-quality dimension asks whether all relevant records and fields are present?

A. Completeness

B. Validity

C. Uniqueness

D. Timeliness

Correct Answer: A. Completeness

Question 4

A correction to an erroneous ledger entry should normally:

A. Overwrite the original without explanation

B. Be deleted after the return is filed

C. Preserve the original information and document the correction

D. Be made only in a personal spreadsheet

Correct Answer: C. Preserve the original information and document the correction

Question 5

Which practice best supports electronic record integrity?

A. Allowing all staff full editing rights

B. Disabling audit logs

C. Using role-based access and version control

D. Saving every document under the same filename

Correct Answer: C. Using role-based access and version control

Question 6

A mixed-use vehicle expense generally requires:

A. Automatic full deduction

B. A documented and reasonable allocation

C. Automatic exclusion of the entire cost

D. No supporting evidence if paid by bank transfer

Correct Answer: B. A documented and reasonable allocation

Question 7

What does least privilege mean?

A. Users receive the minimum access necessary for their roles

B. Only senior managers may use accounting software

C. All records must be publicly accessible

D. Passwords should be shared within a team

Correct Answer: A. Users receive the minimum access necessary for their roles

Question 8

Which statement about AI-generated tax classifications is correct?

A. They are evidence without further review

B. They eliminate the need to retain original documents

C. They require appropriate human validation and governance

D. They may be entered into any public tool without risk

Correct Answer: C. They require appropriate human validation and governance

Question 9

A legal hold generally requires an organisation to:

A. Destroy records immediately

B. Preserve relevant records despite ordinary disposal schedules

C. Remove all access logs

D. Replace source documents with summaries

Correct Answer: B. Preserve relevant records despite ordinary disposal schedules

Question 10

Which item most directly demonstrates the reproducibility of a tax calculation?

A. A final number without workings

B. A working paper retaining inputs, assumptions, formulas, and review evidence

C. An unsigned advertising brochure

D. A blank tax return form

Correct Answer: B. A working paper retaining inputs, assumptions, formulas, and review evidence