Study guide
Data Quality, Integrity, Access, and Taxonomy
CAMS Study Guide Exam
Track your progress as you read
Finished sections get a check mark.
1 Chapter 1: CAMS Orientation and Exam Facts
Current CAMS Exam Format Eligibility Credits and ACAMS Membership Package Pricing and Study Materials Pearson VUE Scheduling Window Passing Score and Result Reporting Exam Non-Disclosure and Practice Ethics
2 Chapter 2: Financial Crime Risks and Methods
AML, CFT, Sanctions, Fraud, ABC, and Tax Evasion Financial Systems and AFC Consequences Predicate Crimes and Financial Crime Exposure Sector, Product, Customer, Jurisdiction, and Channel Risk Banking Products and High-Risk Sectors Politically Exposed Persons and High-Risk Customers
3 Chapter 3: High-Risk Sectors and Structures
MSBs, PSPs, and E-commerce Platforms Insurance Product Risk VASPs, Cryptoassets, and Related Products Gaming, Gambling, and Real Estate Gatekeepers, Trusts, and Company Service Providers Accountants and Other High-Risk Structures
4 Chapter 4: Global AFC Frameworks
FATF Role and Recommendations United Nations Sanctions Regimes Regulators, Law Enforcement, and FIUs Public Sector, Private Sector, and NGO Objectives Cross-Border Cooperation Authority Reports, Trends, and Typologies
5 Chapter 5: Governance Regulations and Information Sharing
Risk-Based Regulatory Expectations Regulatory Examinations and Enforcement Public-Private Partnerships Information Sharing Mechanisms De-Risking and Financial Inclusion RFIs, Tipping Off, and Customer Communication
6 Chapter 6: AFC Compliance Program Design
Effective Program Pillars Three Lines of Defense and MLRO Role Risk Appetite Statement Enterprise Risk Assessment Risk-Based Controls Policies Standards and Procedures
7 Chapter 7: Program Governance and Operating Model
Horizon Scanning and Regulatory Guidance Governance Committees and Terms of Reference AFC Team Structure and Responsibilities Front Office Interaction Supporting Functions and Risk Partners Training, Awareness, and Staff Risk Culture
8 Chapter 8: Customer Lifecycle Controls
Control Tools Across the Customer Lifecycle Data Quality, Integrity, Access, and Taxonomy Customer Experience and Friction Digital Onboarding and Electronic KYC (e-KYC) External Data Sources and Beneficial Ownership Customer Screening and Sanctions Lists
9 Chapter 9: Ongoing Controls and Monitoring
Periodic Refresh and Perpetual KYC Adverse Media Screening Batch Screening and Fuzzy Logic Transaction and Payment Screening Rules-Based Transaction Monitoring Model Risk Management and Tuning
10 Chapter 10: Investigations and Case Support
Investigation Triggers and Case Triage Open Source Tools and Investigator Support Escalation and Suspicious Activity Decisions Documentation and Evidence Quality Law Enforcement and FIU Requests Post-Investigation Feedback Loops
11 Chapter 11: Technology and Emerging AFC Practice
AI and Machine Learning in AFC Controls Blockchain Transaction Screening Automation to Support Investigators Scenario Coverage and Threshold Setting Data Privacy, Cybersecurity, and Other Risk Functions Technology Governance for AFC Tools
12 Chapter 12: Final CAMS Review and Exam Day
Domain-Weighted Study Allocation Simulator Practice Without Actual Exam Claims Pearson VUE Identification and Test Rules Time Management and Guessing Strategy Retake Extension and Appeal Boundaries Recertification and Ongoing Membership
Data Quality, Integrity, Access, and Taxonomy
Key Takeaways Explain This Topic AI
AML systems are only as good as their data: garbage-in produces missed alerts (false negatives) and noise (false positives).
Data quality has measurable dimensions — completeness, accuracy, consistency, timeliness, uniqueness, and validity.
A common taxonomy and a single source of truth (golden record) prevent duplicate customers and fragmented monitoring.
Access controls and data lineage protect confidentiality and create the audit trail regulators and auditors expect.
Last updated: June 2026
Data Quality, Integrity, Access, and Taxonomy
Every sanctions filter, transaction-monitoring scenario, and risk model consumes data. If that data is wrong, incomplete, or fragmented, the control fails silently — the system reports 'clear' while the risk is real. CAMS frames this as the principle that data quality is a control , not a back-office hygiene task.
The six data-quality dimensions
- Dimension Definition AML failure if missing
- Completeness All required fields populated Beneficial owner blank → screening gap
- Accuracy Values reflect reality Wrong date of birth → sanctions match missed
- Consistency Same value across systems Two spellings of one name → fragmented monitoring
- Timeliness Data current and refreshed Stale address → outdated geographic risk
- Uniqueness One record per real entity Duplicate customers → activity split below thresholds
- Validity Conforms to format/rules Free-text country field → screening can't parse
Worked example. A customer is onboarded twice — once as 'Maria Garcia' and once as 'M. Garcia-Lopez' — because of a uniqueness failure. Each record shows $4,000 monthly wires, below a $5,000 monitoring threshold, so no alert fires. Aggregated, the real customer moves $8,000 monthly: classic structuring hidden by poor data. The control did not fail; the data did. The exam answer is entity resolution / deduplication into a golden record , not a new monitoring rule.
Taxonomy and the golden record
A data taxonomy is the agreed classification of customers, products, jurisdictions, and risk categories. Without it, one team's 'PEP' is another's 'high-profile client', and reporting cannot roll up. The golden record (master Customer Information File) is the single, reconciled source of truth that screening and monitoring read from. Reference data — sanctions lists, country-risk ratings, high-risk-business codes — must be versioned so an alert can be reproduced against the list as it stood that day.
Integrity, access, and lineage
Integrity: data must not be altered improperly; changes are logged with who/when/why.
Access control: apply least-privilege and need-to-know. SAR data and investigation files are highly restricted — over-broad access risks tipping off and breaches of confidentiality.
Data lineage: the traceable path from source system → transformation → alert. Examiners and auditors expect you to reproduce why an alert fired and prove the underlying data.
Retention: records (CDD, transactions, SARs) are generally retained for five years under most BSA/FATF regimes; the exam expects you to know that retention enables reconstruction of transactions.
Common traps
First, do not confuse false negatives (a real risk the data hid) with false positives (noise from over-matching). Poor data drives both. Second, fuzzy matching tuned too loose floods investigators; tuned too tight misses sanctioned parties — tuning is a documented, governed decision. Third, adding more rules to a system fed by bad data multiplies the noise; the disciplined fix is upstream data remediation. Fourth, broad data access is not 'efficiency' — uncontrolled access to SAR and investigation data is a reportable confidentiality failure.
The CAMS-correct answer treats data as a governed asset with owners, quality metrics, lineage, and least-privilege access.
Data governance roles and metrics
Good data does not happen by accident; it is governed. A data owner (a business executive) is accountable for the meaning and quality of a data domain, while a data steward operationally maintains it. Data quality metrics — completeness rates, match rates, exception counts — are tracked and reported like any other control metric, with thresholds and escalation. When a regulator asks 'how do you know your customer data is accurate?' the credible answer is a measured, governed program, not an assertion. CAMS scenario items reward identifying the owner and the metric , not just naming the defect.
Reference data and reproducibility
Screening and monitoring depend on reference data : sanctions lists, PEP databases, high-risk country ratings, and high-risk-business codes. This data changes constantly, so it must be versioned and dated . If an alert fired on 1 March against the OFAC list as it stood that day, an investigator must be able to reproduce the match against that exact list version months later. Loading a new list silently over the old one destroys reproducibility and undermines an examiner's ability to test the control — a frequently overlooked defect.
Structured versus unstructured data
Monitoring systems handle structured data (amounts, dates, account numbers in fixed fields) well, but much AML-relevant information arrives as unstructured data — free-text payment messages (the SWIFT MT103 remittance field), adverse-media articles, and scanned documents. A payment whose structured fields look benign may carry a sanctioned port name or a shell-company reference buried in free text. Controls increasingly parse unstructured fields, but free-text entry at onboarding (a country typed instead of selected from a list) is a validity defect that defeats automated screening.
A worked data-lineage scenario
An examiner asks why a $50,000 cash deposit never alerted. Investigation shows the branch system recorded it as a check, not cash, so the cash-structuring scenario never evaluated it. The defect is accuracy at the source , and the fix is upstream input controls plus reconciliation — not a new scenario. Data lineage is what lets the institution trace the alert (or its absence) back to the originating field and prove the root cause.
The exam-correct conclusion: when controls miss, interrogate the data feeding them before adding rules, because new rules built on bad data simply multiply false positives while leaving the true risk invisible.