--- title: "Common Data Quality Use Cases and Where Governance Fits" url: "https://atlan.com/know/data-governance/what-are-common-data-quality-use-cases-and-where-does-governance-fit/" description: "Common data quality use cases, from customer matching to AI training data, and where governance fits: definitions, owners, thresholds and regulation." author: "Emily Winks" author_role: "Data Governance Expert" published: "2026-10-21" updated: "2026-10-21T00:00:00.000Z" --- > Atlan is hosting Context Conference, bringing together the leaders and builders at the frontier of giving AI the context it needs to understand their business. It runs online on October 28, 2026, from 11:00 AM to 2:00 PM ET. Atlan co-founder Prukalpa Sankar opens and closes the day. Leaders from AstraZeneca, BNY and Verizon share why they invest in context and what they get from it. Registrants get early access to The AI Context Gap, a new study from MIT Technology Review Insights. Register: https://atlan.com/context-conference/ ## What are common data quality use cases and where does governance fit? The most common data quality use cases are customer and master data matching, regulatory and risk reporting, trusted analytics, pipeline monitoring, privacy, migration, and data for AI. Governance fits above every one of them: it decides what "good enough" means for each use case, who owns the data, and which failures trigger action. Data quality tools find and measure problems, and governance decides which problems matter and who fixes them. ## Key takeaways - A data quality use case is a business job that fails when the data behind it is wrong, late, or missing. - A governed rule names five things: the field, the dimension it tests, the rule, the owner, and the response when it fails. - Lineage, ownership, and a business glossary turn a quality alert into a task with an owner and a known blast radius. - Start with the use case where bad data already has a price. --- The most common data quality use cases are customer and master data matching, regulatory and risk reporting, trusted analytics, pipeline monitoring, privacy, migration, and [data for AI](https://atlan.com/know/data-for-ai/data-quality-for-ai-agent/). Governance fits above every one of them: it decides what "good enough" means for each use case, who owns the data, and which failures trigger action. Data quality tools find and measure problems, and governance decides which problems matter and who fixes them. --- ## What are the most common data quality use cases? A data quality use case is a business job that fails when the data behind it is wrong, late, or missing. These eight come up most often. | Use case | What breaks when quality slips | Typical example | |---|---|---| | [Customer and master data matching](https://atlan.com/data-catalog-vs-master-data-management/) | Duplicate records split one customer into three | "J. Smith" and "John Smith" in the CRM and support desk | | Regulatory and risk reporting | Filed totals that do not reconcile | Bank risk reports under BCBS 239 | | [Trusted analytics and metrics](https://atlan.com/know/ai-agent/context-layer-for-data-analytics-teams/) | Two dashboards show two revenue numbers | [Finance close](https://atlan.com/know/ai-agent/ai-agents-for-finance/), board packs | | [Pipeline monitoring](https://atlan.com/know/ai-agent/data-for-ai/data-observability-for-ai-pipelines/) | Late feeds, [schema changes](https://atlan.com/know/ai-agent/data-for-ai/data-contracts-for-ai/), sudden null spikes | A daily sales feed that misses its morning deadline | | [Privacy and personal data](https://atlan.com/know/ai-agent/data-for-ai/how-to-handle-pii-in-ai-pipelines/) | Stale addresses, wrong consent flags | [GDPR's accuracy principle](https://atlan.com/know/ai-agent/gdpr-compliance-for-ai-agents/) | | Migration and consolidation | Records dropped or mis-mapped in a move | Merging two ERPs after an acquisition | | Reference and supplier data | Codes that differ by region or product line | Insurance underwriting codes, supplier IDs | | [AI training and agent data](https://atlan.com/know/how-to-ensure-llm-training-data-quality/) | [Models trained on biased sets](https://atlan.com/know/training-data-bias-detection-methods/), [agents answering from stale tables](https://atlan.com/know/llm-knowledge-base-staleness/) | [High-risk AI under the EU AI Act](https://atlan.com/know/ai-readiness/the-eu-ai-act-summary/) | The everyday version of these problems is mundane. A [2019 DATAVERSITY review of governance and quality use cases](https://www.dataversity.net/articles/data-governance-and-data-quality-use-cases/) gives erroneous addresses, incomplete phone numbers, and missing form fields as common examples of errors that data quality tools catch. The scale is not mundane. In a [2017 Harvard Business Review study](https://hbr.org/2017/09/only-3-of-companies-data-meets-basic-quality-standards) by Tadhg Nagle, Thomas Redman, and David Sammon, only 3% of the companies' data met basic quality standards. --- ## Where does data governance fit in data quality? Governance is the set of decisions that tell a data quality program what to check, how strictly, and who answers when a check fails. The DATAVERSITY review puts the relationship plainly: ["Data Governance provides the framework for managing Data Quality."](https://www.dataversity.net/articles/data-governance-and-data-quality-use-cases/) Quality work without that framework produces alerts that nobody owns. Four decisions belong to governance, and each one shapes a quality rule. | Decision | Governance answers | The data quality practice then | |---|---|---| | Definition | What does "active customer" or "net revenue" mean? | Writes a validity rule against that definition | | Ownership | Who is accountable for this table or field? | Routes every failed check to that person | | Threshold | How much error can this use case tolerate? | Sets the pass, warn, and fail levels | | Priority | Which data is critical, and which can wait? | Monitors critical data first and samples the rest | Thresholds show why the split matters. A marketing list can live with some missing phone numbers. A regulatory capital report cannot live with one missing exposure. The check is identical. The tolerance is a business decision, and governance is where that decision gets made, written down, and approved. Governance also makes a failure fixable. A null spike is noise until someone can see which source feeds it, what depends on it, and who signed off on the definition. [Lineage](https://atlan.com/know/ai-agent/data-for-ai/data-lineage-for-ai/), ownership, and a business glossary turn a quality alert into a task with an owner and a known blast radius. The reverse also holds. If no check proves that "customer email" is valid and current, the glossary entry describing it is a claim nobody has tested. The AI Context Stack A short brief from Atlan on the context AI systems need from your data estate. Get the AI Context Stack --- ## What does a governed data quality rule look like? A governed rule names five things: the field, the dimension it tests, the rule, the owner, and the response when it fails. Writing all five down separates a governed check from an engineer's private test. | Field | Dimension | Rule | Owner | Response on failure | |---|---|---|---|---| | customer_id | Uniqueness | No duplicate IDs across CRM and billing | Customer data steward | Block the merge job, open a ticket | | email | Validity | Valid syntax for every opted-in contact | Marketing operations | Suppress from the next send | | order_status | Validity | Only approved status values | Order management lead | Quarantine rows, alert the pipeline team | | daily_sales feed | Timeliness | Lands before the agreed morning cutoff | Data engineering on-call | Flag dashboards as stale | | exposure_amount | Completeness | No nulls in any material risk position | Risk data owner | Hold the regulatory report | The rules above test uniqueness, validity, timeliness, and completeness. Regulators name the dimensions they care about: [BCBS 239](https://www.bis.org/publ/bcbs239.pdf) sets principles for the accuracy, completeness, and timeliness of bank risk data, and [Article 10 of the EU AI Act](https://artificialintelligenceact.eu/article/10/) asks for training data that is free of errors and complete. Each use case leans on a different few. Customer matching depends on uniqueness. Pipeline monitoring depends on timeliness. Risk reporting needs completeness and accuracy together. When a rule fails, a governed response follows a fixed path: 1. Detect the failure and record the rule that caught it. 2. Route it to the named owner. 3. Use [lineage](https://atlan.com/know/ai-readiness/ai-ready-data-lineage/) to find every dashboard, report, and model downstream. 4. Warn those consumers, or mark the asset as untrusted. 5. [Fix the cause at the source](https://atlan.com/know/mcp/data-lineage-rca-with-mcp/) and re-run the check. 6. Log the incident so repeat failures show up in the owner's record. Steps 2, 4, and 6 are governance. Without them, the same broken field fails every Monday and gets patched by whoever notices first. --- ## How do regulations make data quality a governance requirement? Several regulations write data quality into law and hold someone accountable for it, which is why compliance often gets budget first. [Privacy law](https://atlan.com/know/data-privacy-for-ai-agents/) sets an accuracy duty on personal data. [Article 5(1)(d) of the GDPR](https://gdpr-info.eu/art-5-gdpr/) requires personal data to be accurate and, where necessary, kept up to date, and requires inaccurate data to be erased or corrected without delay. A stale address on file is a compliance finding. [Banking regulation](https://atlan.com/know/context-layer-for-financial-services/) goes further. The Basel Committee published its [Principles for effective risk data aggregation and risk reporting](https://www.bis.org/publ/bcbs239.pdf), known as BCBS 239, on 9 January 2013. Principles 3, 4, and 5 cover accuracy and integrity, completeness, and timeliness of risk data. Global systemically important banks designated in 2011 and 2012 had until January 2016 to comply. The Committee's [March 2017 progress report](https://www.bis.org/bcbs/publ/d399.pdf) found that only one of those banks fully complied within the deadline. Its [2015 report](https://www.bis.org/bcbs/publ/d308.pdf) had already flagged banks' dependence on manual processes. AI regulation is the newest entrant. [Article 10 of the EU AI Act](https://artificialintelligenceact.eu/article/10/) requires training, validation, and testing data for [high-risk AI systems](https://atlan.com/know/ai-readiness/ai-risk-management/) to be relevant, sufficiently representative, and, as far as possible, free of errors and complete. It also requires [documented governance practices](https://atlan.com/know/ai-readiness/ai-governance-framework/) covering [data origin](https://atlan.com/know/training-data-lineage-for-llms/), preparation, bias examination, and known gaps. Under the current timeline it applies from 2 December 2027 for Annex III systems. Each rule asks two questions. Is the data good enough for its purpose? Can you show who decided that, and how you know? The first is quality. The second is governance. Context Gap Calculator Estimate the gap between the context your data carries today and what your AI use cases need. Calculate Your Gap --- ## How should a team choose which data quality use cases to govern first? Start with the use case where bad data already has a price. Gartner priced it for AI in a [February 2025 analysis](https://www.gartner.com/en/newsroom/press-releases/2025-02-26-lack-of-ai-ready-data-puts-ai-projects-at-risk), predicting that through 2026 organizations will abandon 60% of AI projects unsupported by [AI-ready data](https://atlan.com/know/ai-agent/data-for-ai/what-makes-data-ai-ready/). Your own number matters more: rework hours, restated reports, or a delayed filing. Tying the program to that cost is also what keeps it alive. In a [February 2024 prediction](https://www.gartner.com/en/newsroom/press-releases/2024-02-28-gartner-predicts-80-percent-of-data-and-analytics-governance-initiatives-will-fail-by-2027-due-to-a-lack-of-a-real-or-manufactured-crisis-), Gartner said 80% of [data and analytics governance initiatives](https://atlan.com/know/ai-agent/ai-agent-governance/why-data-governance-implementations-fail/) will fail by 2027 for lack of a real or manufactured crisis. A quality use case with a visible cost supplies that crisis honestly. A workable sequence: 1. Pick one use case with a business owner, such as regulatory reporting, customer matching, or one executive dashboard. 2. Name its critical data elements, the short list of fields it cannot run without. 3. Have the business owner sign off on what each field means and how much error is tolerable. 4. Write each rule with its field, dimension, rule, owner, and response on failure, and route it to that owner. 5. Show results where people work. A trust signal on the dashboard beats a weekly report nobody opens. 6. Add the next use case once the first one has owners who act on alerts. A frequent failure is the opposite order: profile every table, generate thousands of alerts, and assign none of them. --- ## How does Atlan connect data quality checks to governance? Atlan puts quality rules, their results, and the [governance context](https://atlan.com/know/ai-agent/context-layer-for-data-governance-teams/) around them in the same place, so a failed check arrives with its owner, definition, and [downstream impact](https://atlan.com/know/mcp/mcp-for-data-lineage/) attached. Atlan's [Data Quality Studio, announced on 3 June 2025](https://atlan.com/snowflake-summit-2025-announcements/), lets business users and data engineers define rules together for timeliness, completeness, and compliance thresholds, using plain-language templates or SQL. The rules [run as Snowflake Data Metric Functions inside the customer's own Snowflake account](https://atlan.com/snowflake-summit-2025-announcements/), so checks use existing compute and the data does not leave the warehouse. Results appear as trust signals in Atlan, in BI tools, and in Slack. Because the quality score lives on the asset in the [Enterprise Context Layer](https://atlan.com/data-quality-studio/), it travels with the asset's lineage, glossary terms, and owner. [A failed freshness check](https://atlan.com/know/ai-agent/context-freshness/) on a source table marks the dashboards and AI agents downstream, and agents can read the same score through [Atlan's MCP server](https://atlan.com/know/what-is-atlan-mcp/) before they answer a question from that table. That [shared context](https://atlan.com/know/what-is-the-enterprise-context-layer/) only works when definitions exist at scale. CME Group [cataloged more than 18 million assets and connected 1,300+ glossary terms in its first year on Atlan](https://atlan.com/data-quality-studio/), which gives every quality rule a governed definition to test against. Atlan in Action: Live Context Layer Demos Watch Atlan's context layer in a live demo session. Watch a Live Demo For the two disciplines side by side, read [data quality vs data governance](https://atlan.com/data-quality-vs-data-governance/). For governance programs beyond quality, see [data governance use cases](https://atlan.com/data-governance-use-cases/), and for how quality fits inside a governance framework, see [data quality in data governance](https://atlan.com/data-quality-in-data-governance/). Book a Demo --- ## FAQs about data quality use cases and governance ### Is data quality part of data governance? Yes, in most operating models. Governance sets the definitions, owners, thresholds, and priorities, and data quality management measures and fixes data against them. The two teams often report to different leaders, so shared rules, named owners, and visible results matter more than the org chart. ### What is the difference between data quality and data governance? Data quality describes the condition of data: whether it is complete, accurate, valid, consistent, unique, and timely for a given use. Data governance describes the decisions and accountability around data: who owns it, what it means, who can use it, and what standard it must meet. Quality is measured. Governance is decided, documented, and enforced. ### Who owns data quality in a governance program? A business data owner is accountable for each critical dataset, and a data steward handles day-to-day rules and issue triage. Data engineers build and run the checks and fix pipeline causes. Ownership should be named per table or field, because a failed check routed to a shared team channel is usually a check that nobody fixes. ### What are critical data elements? Critical data elements are the fields a business process, report, or model cannot run correctly without, such as customer ID, exposure amount, or order status. Governance programs name them first so that quality rules, thresholds, and monitoring start where errors cost the most. A first use case usually needs a short list of them. ### How does data quality affect AI agents and models? AI models learn from whatever data they are trained on, and agents answer from whatever tables they can reach. Incomplete, biased, or stale data produces confident wrong answers. From 2 December 2027, Article 10 of the EU AI Act requires documented data governance and quality checks for training data in Annex III high-risk AI systems, and agents need quality scores they can read before they answer. --- ## Sources 1. [Data Governance and Data Quality Use Cases, DATAVERSITY](https://www.dataversity.net/articles/data-governance-and-data-quality-use-cases/) 2. [Only 3% of Companies' Data Meets Basic Quality Standards, Harvard Business Review](https://hbr.org/2017/09/only-3-of-companies-data-meets-basic-quality-standards) 3. [Principles for effective risk data aggregation and risk reporting (BCBS 239), Basel Committee on Banking Supervision](https://www.bis.org/publ/bcbs239.pdf) 4. [Progress in adopting the Principles for effective risk data aggregation and risk reporting, March 2017, Basel Committee on Banking Supervision](https://www.bis.org/bcbs/publ/d399.pdf) 5. [Progress in adopting the Principles for effective risk data aggregation and risk reporting, 2015, Basel Committee on Banking Supervision](https://www.bis.org/bcbs/publ/d308.pdf) 6. [Article 5: Principles relating to processing of personal data, GDPR](https://gdpr-info.eu/art-5-gdpr/) 7. [Article 10: Data and Data Governance, EU Artificial Intelligence Act](https://artificialintelligenceact.eu/article/10/) 8. [Lack of AI-Ready Data Puts AI Projects at Risk, Gartner](https://www.gartner.com/en/newsroom/press-releases/2025-02-26-lack-of-ai-ready-data-puts-ai-projects-at-risk) 9. [Gartner Predicts 80% of D&A Governance Initiatives Will Fail by 2027 Due to a Lack of a Real or Manufactured Crisis, Gartner](https://www.gartner.com/en/newsroom/press-releases/2024-02-28-gartner-predicts-80-percent-of-data-and-analytics-governance-initiatives-will-fail-by-2027-due-to-a-lack-of-a-real-or-manufactured-crisis-)