What Is Data Quality Management? Dimensions, Practices, and Where Marketing Data Breaks

Data quality management is the ongoing practice of keeping data accurate, complete, consistent, timely, unique and valid: the six dimensions the discipline measures against. In practice it combines profiling, cleansing, validation and monitoring, run as a program rather than a project.

Most data quality management assumes the data arrives from a system and is corrected downstream. Marketing data breaks that assumption: campaign names, channel values and creative identifiers are typed by people, at the moment of launch, often by people who do not work for you.

What is data quality management?

The practice of keeping data accurate, complete, consistent, timely, unique and valid across its lifecycle.

The word doing the work is management. Quality is the condition of the data at a point in time; management is the continuing program that measures it, fixes what is wrong, and decides who owns each part.

That distinction matters because quality is not a project with an end. Data degrades as systems change, teams turn over and new sources are added, so a one-off cleanup produces a clean dataset and no capability. The program is the deliverable.

The six dimensions of data quality

The six dimensions are accuracy, completeness, consistency, timeliness, uniqueness and validity.

This is the model the enterprise literature converges on (IBM, “What Is Data Quality Management?”, accessed 2026-09-10), and DAMA International’s framework uses the same six, defining the discipline as ensuring data is “accurate, complete, consistent, timely, valid, and unique” (DAMA International, “DAMA-DMBOK Framework”, DMBOK 2.0, accessed 2026-09-12). Counts differ between publishers. Some collapse validity into accuracy, others add integrity. The properties themselves are stable. A fuller reconciliation of the competing counts sits in the data quality overview.

The standards foundation: Read: what are data standards? — the allowed values these dimensions are measured against.

Why data quality management matters

Poor data quality costs organizations most not through obvious errors but through decisions made confidently on wrong numbers.

An obviously broken dataset gets fixed. Nobody optimizes against a report that is visibly empty. The expensive failures are the ones that look fine: a channel report that is internally consistent and wrong, a cost-per-acquisition figure computed over a population that quietly excludes a third of the campaigns.

Data quality problems blocking analytics, attribution and reporting is the single most prevalent issue in our customer conversations, raised across 96 enterprise accounts. It is almost never reported as a data problem. It arrives as two dashboards disagreeing, or a budget decision someone cannot defend.

The most-cited figure puts the average at $12.9 million per organization per year (Gartner, “How to Improve Your Data Quality”, accessed 2026-09-12). It is worth knowing where that number comes from before quoting it: it originates in Gartner’s 2020 Magic Quadrant research, where 154 reference customers of data-quality vendors were asked what they believed poor data quality was costing them.

That is a self-reported estimate from organizations already shopping for data-quality software — which is to say, organizations that had already concluded they had a problem. The figure is useful as an order of magnitude and misleading as a benchmark, and almost every page that quotes it omits the methodology.

The more defensible framing is proportional. Thomas Redman, writing in MIT Sloan Management Review, put the cost of bad data at 15% to 25% of revenue for most companies, counting the time spent correcting errors, confirming data against other sources, and undoing decisions made on it (MIT Sloan Management Review, “Seizing Opportunity in Data Quality”, Thomas C. Redman, 2017). That is also an estimate rather than a measurement, but it scales to the reader’s own organization in a way a flat dollar figure does not.

The four core practices

Data quality management runs on four practices: profiling, cleansing, validation and monitoring.

• Profiling. Measure what you actually have against each dimension. Distinct-value counts on fields that should be closed sets are the fastest way to surface consistency failures.
• Cleansing. Correct or remove records that fail. Necessary, and the least durable of the four, because it treats symptoms at the far end of the pipeline.
• Validation. Apply rules that reject non-conforming values. Where this runs determines almost everything about whether the program works.
• Monitoring. Track the dimension rates over time so degradation is visible before someone notices it in a report.

That practice set is standard across the discipline (SAS, “Data quality management: what you need to know”, accessed 2026-09-10). The variable is where validation sits, and the standard model puts it after ingestion.

Data quality assurance vs data quality management

Assurance is the checking activity; management is the whole program that decides what to check and who owns the fix.

Assurance sits inside management. A team with assurance and no management has checks that fire, findings nobody owns, and the same failures recurring quarterly.

Where marketing data breaks the standard model

Standard data quality management assumes data arrives from a system; marketing data arrives from people typing into forms and spreadsheets.

“Cleansing marketing data is a treadmill — you are fixing last quarter’s campaigns while this quarter’s are being tagged wrong. The only durable fix is refusing the bad value at entry.” — Bernard Kiyanda, Chief Technology & Product Officer, Claravine Every ranking treatment of this discipline describes a pipeline: data lands, gets profiled, gets cleansed, gets monitored. That model fits transactional and system-generated data well, and it fits marketing data badly, for three reasons.

The error enters before the pipeline. A mistyped channel value is not corrupted in transit; it is created wrong. Profiling will find it, and finding it does not tell you what the value should have been. Nobody can reliably infer that Emial on a specific placement meant email rather than a channel that team genuinely uses.

The volume of authors is high and they rotate. A warehouse table has a handful of writing systems. A campaign taxonomy has every campaign manager, every regional team and every agency, changing composition continuously.

A large share of authors sit outside the perimeter. Cleansing rules, validation jobs and monitoring dashboards all live in systems agencies never touch. Manual entry and copy-paste workflow causing errors and overhead is raised across 66 enterprise accounts, and it is consistently described as a workflow problem rather than a data problem — which is exactly right, and exactly why the data team cannot fix it alone.

How to enforce quality at the point of capture

Enforcing quality at capture means the invalid value cannot be entered, rather than being corrected later.

1. Decide what correct means, per field. Allowed values for closed sets, format patterns for identifiers. Without this there is nothing to enforce.
2. Move validation to the form. The campaign builder, the intake workflow, the setup step: wherever the value is first typed.
3. Reach the external authors. Agencies and partners need the same constrained inputs, in the tools they actually use. A control that covers only internal users covers a minority of the records.
4. Keep downstream checks anyway. Capture-stage enforcement reduces the error rate; it does not make monitoring redundant, and integrations still break.
5. Measure the conformance rate, not the cleanup volume. A falling cleanup volume can mean quality improved or that someone stopped cleaning.

Validating data and enforcing compliance before campaign activation is one of the most common jobs enterprise teams bring us, across 39 accounts, and the framing is consistently pre-launch. The asymmetry is the reason: a rejected value costs somebody seconds, and a wrong value that reaches production costs a quarter of fragmented reporting.

Vanguard’s campaign team described the change when one approach replaced several.

“Before Claravine, people were doing things ten different ways. Now, people have gotten on the bus and are using one consistent approach.” — Mary Daniel, project administrator, Vanguard

How data quality management relates to governance

Governance decides the rules and owners; quality management executes and measures against them.

Governance is the decision-making structure: which fields matter, what the permitted values are, who approves a change, how disputes are resolved. Quality management is the operating discipline that applies those decisions and reports on how well they are holding.

Neither works alone. Governance without quality management produces a policy nobody measures. Quality management without governance produces measurement against rules nobody ratified, which is how a data team ends up defending a threshold in a meeting about a budget. The definitions and owners themselves live in the data dictionary.

Frequently asked questions

Sources

Outbound citations, named and dated:

• IBM, “What Is Data Quality Management?” (accessed 2026-09-10) — the six-dimension model.

• SAS, “Data quality management: what you need to know” (accessed 2026-09-10) — profiling, cleansing, validation and monitoring as the core practice set.

• Gartner, “How to Improve Your Data Quality” (accessed 2026-09-12) — the $12.9M average, published with its methodology in the body.

• MIT Sloan Management Review, “Seizing Opportunity in Data Quality” (Thomas C. Redman, 2017) — the 15–25%-of-revenue framing.

• DAMA International, “DAMA-DMBOK Framework” (DMBOK 2.0, accessed 2026-09-12) — the same six dimensions, from the standards body rather than a vendor.