Data Quality: Where Bad Data Comes From and How to Prevent It (2026)

What Is Data Quality? Four Core Dimensions

No single number captures data quality; you have to read several dimensions together

Data quality really comes down to one question: how usable is your data for the job you actually need it to do? Good data gives your decisions solid ground to stand on, while poor data quietly steers them off course. In practice, we look at it across four dimensions:

1. Accuracy

How faithfully the data mirrors the real world. If a customer really lives at “Ataturk St. No:15,” that is exactly what the record should say. Accuracy problems usually show up as:

• Typographical errors: “Istnabul” instead of “Istanbul”
• Value errors: Entering a unit price of 1000 instead of 100
• Reference errors: An order linked to the wrong customer code
• Measurement errors: Entering grams instead of kilograms

2. Completeness

Whether the fields you actually need are filled in. Gaps in the data both skew your analysis and stall the processes that depend on it. Common completeness problems:

• Empty mandatory fields: A customer record without an email address
• Partial records: An address with a street name but no city/zip code
• Relational gaps: A product card without an assigned category
• Historical gaps: A product without a tracked price change history

3. Consistency

Whether the same fact agrees with itself across your systems. When one record reads two different ways in two places, the operational fallout is bigger than it looks:

• Format inconsistency: “TR-123456” in one system and “123456” in another
• Value inconsistency: A customer marked as “Active” in the CRM but “Inactive” in the ERP
• Temporal inconsistency: An order date occurring after the shipment date
• Calculation inconsistency: Different total revenue figures across various reports

4. Timeliness

Whether the data is current and within reach the moment you need it. A decision made on stale data is a wrong decision from the start:

• Delayed updates: A price increase not reflected in the system
• Delayed integration: Order information taking until the next day to sync to the ERP
• Unarchived data: 10-year-old inactive records appearing as active
• Delayed validation: Address changes discovered months after the fact

Where Does Bad Data Come From? Root Cause Analysis

Any cleanup done without understanding the root cause is only temporary

Almost every data quality issue traces back to one of three sources: people, systems, or processes. Each calls for a different kind of fix, and trying to solve one with another’s playbook just wastes time.

Human-Induced Errors

Anywhere data is keyed in by hand is also a place it can go wrong:

Data Entry Errors

• Typing errors: Typing “1243” instead of “1234”
• Copy-paste errors: Copying the wrong cell
• Format errors: Entering the wrong date format
• Unit errors: Entering quantity instead of weight

Lack of Knowledge

• Lack of training: Not understanding the meaning of a field
• Lack of procedures: Not knowing the correct entry method
• Lack of reference: Not knowing valid values

Intentional Errors

• Time pressure: Skipping mandatory fields for speed
• Bypassing the system: Entering fake values to pass validation
• Lazy entry: Using placeholder values like “X” or “.”

System-Induced Errors

Errors that originate in the technical infrastructure and the software itself:

Integration Errors

• Mapping errors: Incorrect field mapping
• Transformation losses: Data loss during character set conversion
• Synchronization errors: Timing-related incompatibilities
• API errors: Partial data transfer after a timeout

Software Bugs

• Bugs: Calculation or saving errors
• Default values: Incorrectly assigned default values
• Rounding errors: Inconsistencies caused by rounding

Process-Induced Errors

Issues that come from how the business is run and organized:

Design Flaws

• Lack of validation: Fields without entry controls
• Lack of standards: Undefined data formats
• Lack of documentation: Absence of a data dictionary

Management Deficiencies

• Unowned data: Fields without an assigned data steward
• Unmonitored quality: Lack of metrics and reporting
• Lack of audits: Failure to perform periodic checks

Validation Rules and Data Entry Controls

Validation done right stops the error at the door, not after it’s inside

Validation rules are your first line of defense against bad data ever reaching the database. They work at three levels:

Level 1: Format Validation

Checking the technical shape of the data:

• Email format: Checking for the “xxx@domain.com” structure
• Phone format: “+90 5XX XXX XX XX” structure
• IBAN format: Country code + check digit + BBAN
• Date format: DD.MM.YYYY or YYYY-MM-DD
• Tax ID: 10 or 11-digit numeric

Level 2: Business Rule Validation

Checking that the data obeys business logic:

• Range checks: Unit price > 0, stock >= 0
• List checks: Is the country code in the valid list?
• Reference checks: Does the customer code exist in the system?
• Logic checks: Discount rate <= 100%

Level 3: Cross-Field Validation

Checking how fields relate to one another:

• Date relationships: Order date <= Shipment date
• Quantity relationships: Shipped quantity <= Ordered quantity
• Amount relationships: Total = Quantity x Unit Price
• Code relationships: Compatibility of city code + district code

Validation Strategies

Inline Validation

Feedback the instant the user types. Because it catches the error at its source, this is the most effective of the four approaches.

On-Save Validation

Checks every field when the save button is pressed, so the user sees all the problems at once instead of one at a time.

Batch Validation

Kicks in during bulk uploads and integrations, flagging the bad rows so someone can go back and fix them.

Scheduled Validation

Periodic sweeps of the data you already hold. It’s the only way to catch records that have quietly rotted over time.

Data Stewardship: Data Governance Model

Data quality isn’t a project you finish and forget; it’s something you manage continuously. The data stewardship model is what keeps that going.

What Is a Data Steward?

A data steward is the person from a business unit who owns the quality of a specific data domain. Note the emphasis: this sits with the business, not with IT. Their responsibilities typically include:

• Defining and documenting data standards
• Setting quality rules
• Investigating and resolving data issues
• Approving requests for new records
• Monitoring periodic quality reports

Data Steward Assignment Examples

• Customer master: Sales Manager or CRM Manager
• Supplier master: Purchasing Manager
• Product master: Product Manager or R&D Manager
• Financial data: Finance Manager or Chief Accountant
• Employee data: HR Manager

Data Governance Process

1. Definition

• Create a data dictionary (definition, format, and owner for every field)
• Define quality rules
• Determine measurement metrics

2. Measurement

• Calculate automated quality scores
• Visualize on dashboards
• Perform trend analysis

3. Monitoring

• Define threshold values (e.g., accuracy > 98%)
• Set up alerts for deviations
• Initiate corrective actions

4. Improvement

• Perform root cause analysis
• Implement process improvements
• Update training materials

If you’d like to see how data governance plays out in your own industry, take a look at our sector-based solutions.

Field Example: Manufacturing Firm Case Study

Real Case (Brand-Neutral)

Situation

An electronics manufacturer with 85 employees. Before migrating to a new ERP, we measured their data quality, and the picture wasn’t pretty: 18% duplication in the customer master, 12% format inconsistency in product codes, and 8% missing components across the BOMs. The plan on the table was to move to the ERP without fixing any of it first.

Steps Taken

1. Weeks 1-2: Data profiling and quality measurement. We assessed the current state across the four core dimensions and built an issue inventory for each data category.
2. Weeks 3-4: Data steward assignments. The Sales Manager took customer data, the Production Planning Chief took product data, and the Purchasing Manager took supplier data.
3. Weeks 5-8: Cleaning and standardization. Duplicate records were merged, format standards were defined, and missing BOM components were verified with the production team.
4. Weeks 9-10: Integration of validation rules into the system. 47 format, 23 business rule, and 12 cross-field validations were defined in the new ERP.
5. Weeks 11-12: A quality monitoring dashboard and periodic reporting mechanism were established, and weekly quality scorecard meetings began.

Result (Representative)

• Customer duplication rate: 18% -> 0.5% (post-migration)
• Product code format consistency: 88% -> 99.2%
• BOM completeness rate: 92% -> 99.8%
• ERP migration duration: Completed in 4.5 months instead of the planned 6 months
• Post-go-live data-related support requests: 60% below industry average

7 Most Common Data Quality Mistakes

1. Viewing Data Quality as an IT Problem

Data quality is a business problem, not a technical one. IT hands you the tools, but the decisions and the ownership belong in the business units. Your data stewards should come from Sales, Finance, or Operations, not from IT.

2. Performing One-Time Cleaning

The “we cleaned it once, we’re done” mindset. Data gets dirty continuously, and without validation rules, monitoring, and periodic audits, any cleanup you do is only temporary.

3. Accepting Data Without Validation

Letting users enter whatever they like on the “we’ll fix it later” assumption. Stopping bad data at the source is far cheaper and far more effective than scrubbing it afterward.

4. Not Documenting Data Standards

Everyone inventing their own format because no one can answer “how am I supposed to enter this?” Your data dictionary and entry standards need to be written down and easy to find.

5. Not Measuring Quality Metrics

Declaring “our data is fine” without ever measuring it. You can’t manage what you don’t measure. Define accuracy, completeness, consistency, and timeliness metrics for every data category.

6. Intervening Only for Critical Errors

Waving off the small stuff. Little errors pile up, and eventually the pile becomes a crisis. Proactive monitoring and early intervention are what keep that from happening.

7. Skipping User Training

The system is live and the validations are in place, but no one has told users why or how to enter clean data. Without training, behavior doesn’t change.

A systematic approach heads errors off before they start

Data Quality Success Metrics

Key metrics you should measure to manage data quality (representative target values):

Track these metrics weekly or monthly, and always watch the trend over time rather than a single snapshot.

Data Quality Checklist

Frequently Asked Questions (FAQ)