Workforce Data Architecture, Quality, and Governance

Introduction

Workforce metrics are only as credible as the data architecture that produces them. A turnover rate, labor-cost ratio, time-to-fill measure, or learning-completion percentage may appear objective, yet each depends on definitions, source systems, data transformations, ownership decisions, and controls. If employee status is defined differently in the human resources information system (HRIS), payroll, and finance system, the resulting metric may be numerically precise but conceptually invalid.

Workforce data architecture is the organized structure through which people-related data is captured, stored, integrated, transformed, accessed, protected, and used. Workforce data governance is the system of decision rights, accountability, standards, policies, and controls that ensures the data is reliable, lawful, secure, and fit for purpose. Together, architecture and governance form the infrastructure of evidence-based HR management.

For graduate HR professionals, the challenge is not merely learning how to extract a report. It is learning how to evaluate the chain of evidence behind a report and design controls that make workforce information trustworthy. This requires technical literacy, managerial judgment, ethical awareness, and an understanding of privacy and regulatory obligations.

Learning Objectives

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

1. Map workforce data across HRIS, payroll, applicant tracking, learning, engagement, finance, and operational systems.
2. Distinguish systems of record, systems of reference, analytical stores, and reporting layers.
3. Evaluate data quality using dimensions such as accuracy, completeness, consistency, timeliness, uniqueness, validity, and integrity.
4. Design master-data structures and resolve conflicting definitions across systems.
5. Assign data ownership, stewardship, and accountability using governance frameworks.
6. Explain metadata, lineage, access controls, privacy, security, retention, auditability, and regulatory compliance.
7. Apply governance principles to a practical workforce analytics case.
8. Critically assess the limitations and ethical risks of workforce data and AI-enabled analytics.

1. The Workforce Data Ecosystem

A workforce data ecosystem consists of interconnected systems that capture different stages of the employment relationship. No single system normally contains a complete and authoritative view of the workforce.

These systems differ in purpose, granularity, timing, identifiers, and definitions. Payroll may represent a worker through a payroll number, while the HRIS uses an employee identifier and the ATS uses a candidate identifier. A person can therefore appear as a candidate, employee, contingent worker, alumnus, or former employee, sometimes simultaneously.

Systems of record and systems of reference

A system of record is the designated authoritative source for a particular data attribute or transaction. For example, the HRIS may be the system of record for employment status, while payroll is the system of record for gross pay actually processed. A system of reference provides useful information but is not authoritative for that attribute.

Authority is attribute-specific. It is usually incorrect to declare an entire application the authoritative source for all workforce data. The payroll system may be authoritative for paid earnings but not for organizational hierarchy; the HRIS may be authoritative for job title but not for actual hours paid.

A layered architecture

A mature architecture commonly contains five layers:

1. Source layer: operational applications where data is created.
2. Integration layer: application programming interfaces, files, event streams, or integration platforms that move data.
3. Storage layer: databases, warehouses, lakehouses, or governed data marts.
4. Semantic layer: standardized definitions, dimensions, measures, and business rules.
5. Consumption layer: dashboards, reports, statistical models, self-service tools, and regulatory submissions.

The semantic layer is especially important. It converts technical fields into shared organizational meanings. For example, “active headcount” should specify whether it includes employees on paid leave, unpaid leave, interns, secondees, and workers with future start dates.

2. Data Mapping and Lineage

Data mapping documents how a field moves from its source to its final use. A useful mapping identifies the source field, target field, transformation rule, data type, owner, frequency, sensitivity classification, and validation control.

Data lineage

Data lineage is the traceable history of data from creation through transformation and use. It answers questions such as:

• Where did this value originate?
• Which transformations changed it?
• Which reports depend on it?
• Who approved the definition?
• What was the value at a particular historical point?

Lineage is essential when a manager challenges a result, when an auditor requests evidence, or when an analyst discovers a sudden change in a trend. Lineage may be documented manually, through metadata repositories, or automatically through data-engineering tools. Automated lineage is valuable but does not replace human interpretation of business rules.

Grain and keys

Every dataset should have an explicitly defined grain: the level represented by one row. Examples include one row per employee, one row per employee per pay period, one row per requisition, or one row per survey response. Joining datasets with different grains can create duplication and materially distort metrics.

A primary key uniquely identifies a record within a dataset. A foreign key links that record to another entity. Workforce architectures often use:

• Employee ID for the employment relationship;
• Person ID for an individual across multiple employment periods;
• Position ID for an authorized organizational position;
• Requisition ID for a recruitment request;
• Cost-center code for financial responsibility;
• Course or learning-object ID for a learning activity.

A robust design distinguishes a person from an employment relationship. One individual may have multiple employments, assignments, or contracts. Treating person ID and employee ID as interchangeable can lead to double counting or inappropriate historical overwriting.

3. Data Quality as a Management Discipline

Data quality is fitness for a defined purpose. A field can be adequate for one use and inadequate for another. A job title may be sufficient for an employee directory but too inconsistent for occupational analysis. Quality must therefore be assessed in relation to a business question, population, time period, and decision consequence.

The principal dimensions are:

• Accuracy: The value correctly represents the real-world object or event.
• Completeness: Required values are present.
• Consistency: The same concept is represented coherently across records and systems.
• Timeliness: Data is available within the period required for use.
• Validity: Values conform to permitted formats, ranges, and rules.
• Uniqueness: Duplicate records are absent or intentionally identified.
• Integrity: Relationships between records remain valid.
• Accessibility: Authorized users can obtain the data when needed.

A simple completeness rate can be calculated as follows:

The numerator counts populated required fields, and the denominator counts all required field instances. This measure assumes that a null value is the relevant indicator of missingness and does not prove that populated values are accurate.

For example, if 9,600 of 10,000 employee records contain a valid manager ID, completeness is 96%. The measure does not establish whether the 9,600 manager IDs refer to current managers; that is a validity and integrity question.

A data-quality score should not conceal serious defects through simple averaging. A dataset with high completeness but invalid compensation values may be unusable. Critical fields should therefore have threshold rules and severity classifications.

Data-quality controls

Controls may be preventive, detective, or corrective.

• Preventive controls stop invalid data at entry, such as permitted-value lists for employment status.
• Detective controls identify defects after entry, such as a report of employees without valid cost centers.
• Corrective controls resolve defects, such as an approved data-cleansing workflow.

Examples of validation rules include:

• hire date cannot be later than termination date;
• birth date must satisfy an approved privacy and plausibility policy;
• an active employee must have an employing entity and organizational assignment;
• a requisition closure date cannot precede its opening date;
• a learning completion cannot precede enrolment unless the system supports prior credit;
• a payroll transaction must reconcile to an approved payroll period;
• each cost center must map to an active finance hierarchy.

Quality management should use issue logs that record the defect, affected population, root cause, owner, priority, remediation date, and evidence of closure. Repeated defects usually indicate a process or system-design problem rather than an individual data-entry mistake.

4. Master Data and Common Definitions

Master data describes relatively stable entities shared across processes. Workforce master data commonly includes person, employment, position, job, organization, location, legal entity, cost center, manager, and calendar dimensions.

Reference data consists of controlled values used to classify data, such as employment-status codes, country codes, job families, absence categories, and education levels. Reference data should have an approved owner, description, effective dates, and retirement procedure.

The importance of effective dating

Organizational structures and job assignments change over time. Overwriting a department value may make current reporting correct while corrupting historical analysis. Effective-dated records preserve the state of an entity during a defined period. A historical report should use the hierarchy applicable at the time of the event, unless the analytical purpose explicitly requires restating history under the current structure.

Master-data governance

A master-data program should establish:

1. canonical entities and identifiers;
2. authoritative sources for each attribute;
3. naming and coding standards;
4. duplicate-resolution rules;
5. effective-dating conventions;
6. synchronization frequency and failure handling;
7. stewardship responsibilities;
8. change-approval procedures.

A common organizational taxonomy is a controlled classification of business units, departments, teams, jobs, grades, locations, and cost centers. Taxonomies should be sufficiently stable for analysis but flexible enough to reflect legitimate organizational change. Excessive local customization prevents comparison; excessive centralization can make the structure irrelevant to operational reality.

5. Governance, Ownership, and Decision Rights

Governance answers who may decide, who must perform, who must be consulted, and who must be informed. It is not identical to data management. Data management performs technical and operational activities; governance establishes authority, accountability, and standards.

A practical accountability structure may include:

• Data owner: Senior person accountable for the business meaning, risk, and permitted use of a data domain.
• Data steward: Person responsible for definitions, quality monitoring, issue resolution, and day-to-day coordination.
• System owner: Person accountable for the application’s operation, configuration, and availability.
• Data custodian: Technical party responsible for storage, backup, infrastructure, and security implementation.
• Data user: Authorized person who accesses or analyzes data for a legitimate purpose.
• Privacy or compliance officer: Specialist who advises on lawful processing, rights, and regulatory obligations.

A RACI matrix can clarify responsibilities: Responsible performs the activity, Accountable owns the outcome, Consulted provides input, and Informed receives updates. One role may hold several designations, but accountability should not be ambiguous.

Governance bodies often include a data council, domain working groups, information-security representatives, HR leaders, finance, legal, and employee-relations specialists. Their decisions should be recorded in policies, data dictionaries, standards, and exception registers.

6. Metadata and the Semantic Foundation

Metadata is data about data. It may describe a field’s name, definition, data type, permitted values, source, owner, sensitivity, update frequency, lineage, quality threshold, and retention period.

A data dictionary might define “voluntary turnover” as the number of employees whose employment ended during a period for employee-initiated reasons, divided by average headcount for that period. The definition should also specify exclusions, treatment of transfers, population scope, and effective date.

Metadata has several forms:

• Business metadata: meaning, owner, policy, and permitted interpretation.
• Technical metadata: table, field, data type, interface, and storage location.
• Operational metadata: load time, refresh status, record counts, and failures.
• Lineage metadata: upstream sources and downstream dependencies.
• Security metadata: sensitivity classification, access role, and handling requirements.

Without metadata, self-service analytics can produce multiple competing versions of supposedly identical metrics. A catalog and glossary reduce this risk, but only if definitions are governed and maintained.

7. Access Control, Privacy, and Security

Workforce information can include identity data, compensation, health information, performance records, survey responses, disciplinary material, and protected characteristics. Access should follow least privilege: users receive only the access necessary for an authorized role and purpose.

Important controls include:

• role-based access control;
• segregation of duties;
• multi-factor authentication;
• encryption in transit and at rest;
• privileged-access monitoring;
• periodic access recertification;
• secure development and change management;
• backup and recovery testing;
• incident detection and response;
• masking or pseudonymization in analytical environments.

Pseudonymization replaces direct identifiers with coded identifiers while preserving controlled analytical linkage. It reduces exposure but does not necessarily make data anonymous, because re-identification may remain possible when datasets are combined.

Privacy governance begins with purpose limitation and data minimization. Organizations should collect and use data for specified legitimate purposes, avoid unnecessary fields, restrict secondary use, and communicate relevant practices transparently. Sensitive workforce analytics should consider whether individuals could reasonably expect the use, whether the analysis is proportionate, and whether group reporting creates a risk of re-identification.

Small-cell suppression may be appropriate when reporting demographic or engagement results. However, suppression thresholds should be documented and applied consistently. Removing a group from a dashboard can itself create misleading conclusions, so suppressed results should be clearly labeled.

Regulatory compliance

Exact obligations vary by jurisdiction and sector. Common regulatory themes include lawful processing, transparency, access and correction rights, security safeguards, breach response, records management, anti-discrimination, wage and hour requirements, tax reporting, and restrictions on sensitive data. Organizations must identify the jurisdictions in which employees, candidates, and systems are located and obtain qualified legal advice for specific obligations.

Compliance should not be treated as a checklist separate from analytics. A metric may be technically accurate but unlawful to use for a particular employment decision. Automated screening, profiling, monitoring, and inferred attributes require heightened review for fairness, explainability, necessity, and human oversight.

8. Retention, Auditability, and Records Management

Retention is the controlled duration for which information is kept. It should reflect legal requirements, operational necessity, litigation holds, historical value, and privacy risk. Keeping everything forever increases exposure and storage costs; deleting too early can undermine legal rights, financial control, or institutional memory.

A retention schedule should identify the record class, responsible owner, retention trigger, retention period, legal hold procedure, archival status, and authorized disposal method. A trigger may be the end of employment, closure of a recruitment process, completion of a pay year, or resolution of a dispute.

Auditability requires the ability to reconstruct what happened. Useful evidence includes:

• who accessed or changed a record;
• what value changed and when;
• the old and new values;
• which rule or workflow authorized the change;
• which version of a report or model was used;
• how a metric was calculated;
• whether an exception was approved.

Audit logs should themselves be protected from unauthorized alteration. Reconciliation controls are particularly important for payroll and finance. For example, total payroll expense in an analytical warehouse should be reconciled to the controlled payroll or general-ledger totals, with documented explanations for timing and scope differences.

9. A Worked Data-Governance Example

Assume an organization reports monthly headcount. The HRIS shows 12,400 active employees, payroll shows 12,180 paid employees, and finance shows labor costs for 12,050 workers. An executive asks which number is correct.

There is no automatic single answer. The analyst must first define the question:

• For organizational capacity, HRIS active headcount may be appropriate.
• For workers receiving pay in the month, payroll population may be appropriate.
• For labor-cost allocation, finance’s controlled accounting population may be appropriate.

The analyst then documents scope differences. HRIS may include employees on unpaid leave or employees not yet paid in the month. Payroll may exclude unpaid workers and include retroactive adjustments. Finance may use accounting-period rules, exclude some worker categories, or record costs under different entities.

The correct response is not to select the largest number but to publish a metric definition, population rule, period convention, source authority, reconciliation statement, and limitations. If the organization needs a common workforce measure, governance leaders should approve a canonical headcount definition and related variants rather than forcing all purposes into one number.

10. Illustrative Case: Northstar Services

Background

Northstar Services, a fictional multinational customer-support organization, has 18,000 workers across six countries. Its executive committee receives monthly reports on turnover, labor cost, recruitment, and employee engagement. The reports are produced by HR, finance, and operations using separate spreadsheets and extracts.

Problem

The reported annual turnover rate differs by as much as 8 percentage points between HR and finance. Recruitment claims that time-to-fill is improving, while business leaders observe persistent vacancies. Engagement results cannot be reliably compared across regions because response rates and employee populations differ. An internal audit also finds that former employees retain access to a workforce dashboard for several months.

Analysis

A diagnostic review identifies five root causes:

1. HR and finance use different definitions of active employee and termination date.
2. Employee IDs are not consistently used across the HRIS, payroll, and finance system.
3. Regional departments use local cost-center and job-title codes.
4. Survey data is joined to employee data without a documented small-group privacy rule.
5. Dashboard access is granted through broad organizational roles and is not recertified regularly.

The organization’s problem is therefore architectural and governance-related, not simply a reporting error.

Approach and solution

Northstar establishes a workforce data council chaired by the HR operations executive, with finance, information security, legal, employee relations, and regional representatives. The council approves the following design:

• HRIS becomes authoritative for employment status and organizational assignment.
• Payroll remains authoritative for paid earnings and pay-period transactions.
• Finance remains authoritative for ledger amounts and cost-center accounting.
• A person-and-employment master model separates person identity from employment relationships.
• A controlled organization and job taxonomy is introduced with effective dates.
• A shared glossary defines headcount, average headcount, turnover, requisition age, time-to-fill, and engagement response rate.
• Data-quality dashboards monitor missing IDs, invalid manager relationships, unmatched cost centers, and delayed interfaces.
• Sensitive survey results are reported only at approved aggregation levels.
• Role-based access is implemented, with quarterly access recertification and immediate termination of access when employment ends.
• A retention schedule and audit-log standard are adopted.

The project also introduces a reconciliation process. Each monthly report contains a “metric contract” stating the definition, source, extraction date, population, exclusions, transformations, and responsible owner.

Outcome

After two reporting cycles, differences between HR and finance are explained rather than silently mixed. Leadership receives separate measures for active workforce, paid workforce, and labor-cost population. Recruitment measures become more credible because requisition and offer dates are standardized. Survey reporting improves because response rates use a documented denominator and sensitive results are protected.

The organization does not eliminate all disagreement. Instead, it makes disagreements visible, traceable, and governable. This is a more realistic and valuable outcome than assuming that one universal dataset will answer every workforce question.

Learning points

The case demonstrates that data governance is an operating capability. It requires executive sponsorship, technical design, process ownership, privacy controls, and continuous maintenance. It also illustrates that reconciliation is not the same as forcing identical values: legitimate differences can remain when their definitions and purposes are explicit.

11. Technology, Cloud Platforms, and AI

Cloud data warehouses, lakehouses, application programming interfaces, event-driven integration, and metadata platforms have made workforce data more accessible. They can improve scalability, refresh speed, lineage, and self-service analysis. They also increase the risk that sensitive data will be copied into uncontrolled locations.

AI and machine-learning systems may support anomaly detection, record matching, forecasting, classification, and natural-language access to governed data. For example, an anomaly model may flag an unusual increase in termination records or a sudden decline in learning completions. Such tools should augment, not replace, accountable review.

AI governance should address data provenance, model purpose, training-data quality, bias testing, explainability, human oversight, security, monitoring, and appeal mechanisms. A model that predicts attrition may reproduce historical inequities if past promotion or scheduling decisions were biased. Predictive accuracy alone is not sufficient justification for using a model in employment decisions.

Generative AI introduces additional risks. Users may paste confidential workforce records into unapproved tools, receive fabricated explanations, or overlook disclosure and copyright obligations. Approved tools should use access-controlled data, logging, retention restrictions, and clear rules against exposing identifiable employee information unnecessarily.

12. Challenges and Limitations

Several structural challenges persist:

• Legacy systems: Old applications may lack APIs, effective dating, or reliable identifiers.
• Organizational change: Mergers and reorganizations disrupt taxonomies and historical comparability.
• Semantic ambiguity: Terms such as employee, worker, headcount, and turnover have multiple legitimate meanings.
• Data silos: Local ownership can prevent enterprise consistency.
• Quality trade-offs: Immediate operational needs may conflict with ideal cleansing processes.
• Privacy tension: More detailed data may support analysis while increasing individual risk.
• False precision: Dashboards can make uncertain or biased measures appear authoritative.
• Global variation: Legal requirements, language, calendars, and employment categories differ by jurisdiction.

Governance should therefore be risk-based. Critical payroll, identity, legal, and safety data normally requires stronger controls than low-risk aggregate information. Governance should also be proportionate: excessive approval processes can encourage shadow systems and reduce adoption.

13. Best-Practice Implementation Roadmap

A practical implementation sequence is:

1. Identify high-value workforce decisions and associated data products.
2. Inventory systems, datasets, interfaces, owners, and sensitive fields.
3. Define critical data elements and business glossaries.
4. Document lineage, grain, keys, and transformation rules.
5. Assign ownership and stewardship through formal decision rights.
6. Establish quality rules, thresholds, issue management, and reconciliation.
7. Implement identity, role-based access, encryption, logging, and recertification.
8. Create retention, archival, deletion, and legal-hold procedures.
9. Test dashboards and models with representative users and affected groups.
10. Monitor controls continuously and review definitions after organizational or legal change.

Maturity should be evaluated not by the number of tools acquired but by whether decisions are supported by data that is defined, traceable, protected, and appropriately used.

Chapter Summary

Workforce data architecture connects operational systems, integration processes, analytical storage, semantic definitions, and reporting tools. Because each system serves a different purpose, authority must be assigned at the attribute level rather than assumed for an entire application. Data mapping, lineage, explicit grain, and reliable identifiers are foundational to trustworthy metrics.

Data quality includes accuracy, completeness, consistency, timeliness, validity, uniqueness, integrity, and accessibility. Preventive, detective, and corrective controls should be linked to business risk. Master data and reference data create shared organizational language, while effective dating preserves historical meaning.

Governance establishes ownership, stewardship, decision rights, metadata standards, access controls, privacy safeguards, retention rules, and auditability. Least privilege, purpose limitation, data minimization, secure processing, and documented regulatory review are essential when handling workforce information. AI and cloud technologies can improve analysis but require provenance, fairness, human oversight, and strong security. The ultimate goal is not a single universal workforce number; it is a transparent, controlled, and fit-for-purpose evidence system.

Key Terms

Data architecture: The structure through which data is captured, integrated, stored, transformed, accessed, and used.

Data governance: Decision rights, accountability, standards, policies, and controls for managing data.

System of record: The designated authoritative source for a particular data attribute or transaction.

Data lineage: The traceable history of data from origin through transformation and use.

Data grain: The level of detail represented by one row in a dataset.

Primary key: A field or combination of fields that uniquely identifies a record.

Master data: Shared core entities such as people, jobs, organizations, and cost centers.

Reference data: Controlled classification values used to categorize records.

Metadata: Information describing data’s meaning, structure, origin, quality, ownership, and use.

Data quality: Fitness of data for a defined purpose.

Data owner: The person accountable for a data domain’s meaning, risk, and permitted use.

Data steward: The person responsible for day-to-day definitions, quality, and issue coordination.

Least privilege: Providing only the access necessary for an authorized purpose.

Pseudonymization: Replacing direct identifiers with coded identifiers while preserving controlled linkage.

Purpose limitation: Restricting data use to specified and legitimate purposes.

Retention schedule: A controlled specification of how long records are kept and when they are disposed of.

Auditability: The ability to reconstruct actions, changes, decisions, and calculations.

Effective dating: Recording the period during which a value or relationship is valid.

Metric definition: A documented specification of a measure’s population, formula, source, period, and exclusions.

Data custodian: The technical party responsible for storage, infrastructure, and security implementation.

Review Questions

Short Answer Questions

1. Why can a system be authoritative for one workforce attribute but not for another?
2. Define data lineage and explain why it matters in workforce reporting.
3. Distinguish master data from reference data.
4. What is the principle of least privilege?
5. Why is effective dating important for organizational and workforce analysis?

Long Answer Questions

1. Explain the major layers of a workforce data architecture and describe the role of the semantic layer.
2. Discuss the principal dimensions of data quality and explain why quality must be evaluated in relation to purpose.
3. Compare the responsibilities of data owners, data stewards, system owners, custodians, and data users.
4. Develop a governance framework for protecting compensation, engagement, and performance data.
5. Evaluate the opportunities and risks created by cloud platforms and AI in workforce analytics.

Analytical / Application Questions

1. An HRIS reports 5,000 active employees, payroll reports 4,850 paid employees, and finance reports labor costs for 4,700 workers. Design an investigation that determines whether the differences are errors or legitimate scope differences.
2. Design a data map for calculating time-to-fill using an ATS, HRIS, and finance system. Specify the grain, keys, authoritative fields, transformations, and quality controls.
3. A manager requests individual engagement scores for a team of four employees. Analyze the privacy, ethical, and governance issues and recommend an appropriate reporting response.

Multiple Choice Questions with Answers

Question 1

Which system is most appropriately treated as authoritative for earnings actually processed during a pay period?

A. Applicant tracking system

B. Payroll system

C. Learning management system

D. Engagement platform

Correct Answer: B. Payroll system

Question 2

What does data grain describe?

A. The security classification of a dataset

B. The level of detail represented by one row

C. The age of a database

D. The number of users with access

Correct Answer: B. The level of detail represented by one row

Question 3

If 9,600 of 10,000 required employee fields are populated, what is the completeness rate?

A. 9.6%

B. 40%

C. 96%

D. 104.2%

Correct Answer: C. 96%

Question 4

Which practice best preserves historical organizational analysis?

A. Overwriting all prior department values with current values

B. Deleting records after each reorganization

C. Using effective-dated organizational records

D. Allowing each analyst to create a local hierarchy

Correct Answer: C. Using effective-dated organizational records

Question 5

Who is primarily accountable for the business meaning and risk of a data domain?

A. Data owner

B. Data user

C. Database viewer

D. Report consumer

Correct Answer: A. Data owner

Question 6

Which control is preventive?

A. A report listing invalid cost centers after loading

B. A permitted-value list that blocks invalid status codes

C. A post hoc correction of duplicate records

D. An audit investigation after a breach

Correct Answer: B. A permitted-value list that blocks invalid status codes

Question 7

What is the main purpose of a data dictionary?

A. To replace all access controls

B. To document definitions, attributes, and permitted interpretations

C. To calculate payroll taxes automatically

D. To archive obsolete hardware

Correct Answer: B. To document definitions, attributes, and permitted interpretations

Question 8

Which principle limits access to only what a user needs for an authorized role?

A. Data duplication

B. Least privilege

C. Universal access

D. Historical restatement

Correct Answer: B. Least privilege

Question 9

Why is pseudonymized workforce data not necessarily anonymous?

A. It always contains payroll totals

B. It cannot be stored securely

C. Individuals may still be re-identified by combining data

D. It removes all analytical value

Correct Answer: C. Individuals may still be re-identified by combining data

Question 10

A metric contract should normally include which combination?

A. Only the dashboard color and publication date

B. Definition, population, source, transformations, and owner

C. Only the analyst’s personal interpretation

D. Only the database password and server location

Correct Answer: B. Definition, population, source, transformations, and owner