Data Quality: What Businesses Can Learn from Claude Monet

Le Havre, 1872: Claude Monet stands at the window of his hotel room, painting the sunrise over the harbor. What he puts on canvas initially unsettles viewers: blurred cranes, a diffuse orange glow in the haze, boats reduced to vague outlines. A critic later mocked the painting as not even finished, but merely an “impression.” The term stuck. Impression, Sunrise gave an entire artistic movement its name: Impressionism.

There is an intriguing theory about why Monet painted the harbor the way he did: He may have been nearsighted. In that case, the painting did not necessarily show the harbor as it was, but rather as Monet perceived it.

Seen through today’s lens, Monet may have had a classic data quality problem.

Claude Monet’s Faulty Data Pipeline

Let’s stay with the analogy for a moment: Monet’s business process was painting. His input was the reality outside the window, his sensors were his eyes, and his output was the painting.

Between reality and canvas was a processing chain—essentially, a data pipeline. And that pipeline may have contained a defect: The raw data—the harbor of Le Havre, with its cranes, steamships, and commercial activity—arrived blurred, noisy, and color-shifted. Monet turned it into art history.

This diagnosis is not entirely clear-cut. Opinions differ on whether Monet was actually nearsighted. What is certain, however, is that he developed cataracts later in life. Over the years, his paintings became more reddish and yellowish because his lens increasingly filtered out blue tones. After surgery in 1923, Monet is said to have been alarmed by how much his perception had changed. He subsequently reworked or destroyed some of his later works.

That takes the analogy beyond a one-time pipeline issue. It is about a gradual deterioration in data quality without monitoring. The error does not occur suddenly—it develops continuously. Without appropriate measurement, validation, and accountability, it often goes unnoticed for a long time.

Why Does Data Quality Matter to Decision-Makers?

The point is not that Monet needed a data engineer. For art history, the blur was actually a stroke of luck.

For businesses, however, the standard is different. Monet painted a major commercial harbor. Had he been a merchant, he would have needed to read the harbor as a marketplace full of information: Which ships are docked? What goods are being bought? Where are business opportunities emerging today?

If competitors capture this information more quickly or reliably, that creates a disadvantage. While others are already taking action, you may still be searching for the relevant signals.

That is precisely why data is your company’s sensory system.

Just as Monet’s eyes stood between the harbor and the canvas, your data stands between reality and decision-making. You do not perceive markets, clients, competitors, and internal business processes directly. You see them through reports, dashboards, Excel spreadsheets, operational systems, and informal updates.

That is why data quality is not only about technically correct data sets. It is about whether data is fit for purpose: for managing business processes, analyzing client behavior, assessing risk, or making strategic decisions.

Poor data quality can result in numbers that appear plausible but do not provide a sound basis for decision-making. Good data quality, by contrast, creates a shared, trustworthy foundation of facts.

Poor Data Quality: Common Symptoms in Organizations

Many companies are familiar with situations like these:

• Sales and Controlling report different revenue figures. In the monthly meeting, the first discussion is about which number is correct instead of which actions to take and which decisions to make.
• A project’s budget is stored in one system, while actual costs are recorded in another. A reliable view of the margin either does not exist or requires significant manual effort.
• An important early indicator of client churn exists in principle but is spread across multiple Excel files and data sources.
• Three departments answer the question, “How did Q2 go?” with three different figures and interpretations.
• Data repositories contain incomplete or incorrect data, duplicates, and different formats for the same information.

It is like the harbor of Le Havre in the morning mist: You get a picture of reality, but it remains an impression.

Data quality issues are not limited to individual data sets, however. They can arise when data from different sources is consolidated without aligning definitions, metadata, and standards. Manual transfers, missing validation, or unclear ownership can also lead to inconsistencies.

Defining Data Quality: When Is Data Truly Usable?

There is no universal definition of data quality separate from the specific use case. Data quality is measured by whether the data meets the requirements of a particular decision, analysis, or automation process.

The following data quality dimensions are especially relevant:

• Accuracy: Does the data reflect the reality it represents? An inaccurate client address, an incorrect revenue figure, or a misallocated cost center can impair analyses and downstream processes.
• Completeness: Do data sets contain all the information required for their purpose? Missing values can significantly limit the usefulness of reports.
• Consistency: Is information uniform across systems, time periods, and business functions? If a client is represented differently in different data sources, there is no consistent view.
• Validity: Do values comply with agreed formats, permitted value ranges, and business rules?
• Uniqueness: Can an object—such as a client, product, or contract—be clearly identified? Otherwise, duplicates can skew analyses.
• Relevance: Does the available data actually support the specific question and decision at hand?

These data quality dimensions demonstrate that data quality is not purely a technical issue. Accuracy, consistency, and validity must be defined from a business perspective and embedded within the organization.

Measuring Data Quality: From Gut Feel to Clear Metrics

“We have a problem with our data” is an important observation, but it is not yet a robust analysis. Measuring data quality requires clear expectations for critical data sets and traceable metrics.

Data quality measurement therefore starts with specific questions: Which data is especially relevant to a process or decision? What quality requirements apply? And how can you regularly verify that these requirements are being met?

For example, trustworthy client analysis may require client data to be complete, consistent, and unique. Operational management may additionally depend on data timeliness. Requirements vary by use case, data volume, and risk profile.

Data quality should not be assessed just once. Data landscapes are constantly changing: New data sources are added, interfaces are modified, and large volumes of data are processed automatically. It therefore makes sense to measure data quality continuously and make changes visible.

Data Quality Management: Accountability, Standards, and Governance

Data quality management means managing, measuring, and improving data quality systematically. It is not just about cleaning up individual data errors. It is about repeatable processes, clear accountability, and appropriate data quality standards.

Effective data quality management brings together several perspectives:

• Business functions define which data is relevant to their responsibilities and which quality requirements apply.
• Data and IT teams create the technical foundation for integrating, validating, and, where appropriate, automating data from different sources in a controlled way.
• Data governance determines who is accountable for data, which rules apply, and how deviations are handled.
• Appropriate tools can support measurement, monitoring, and data cleansing. However, they do not replace business definitions or clear ownership.

Data governance and data integrity are closely connected. Put simply, data integrity means that data remains complete, accurate, and consistent when it is stored, processed, or transferred between systems. Governance provides the framework for implementing these requirements transparently across the organization.

The goal is not to perfect every piece of information with maximum effort. High data quality means that the quality of the data fits the respective use case and is reliable enough for the decisions that matter.

Data Quality Challenges and the Use of AI

Data quality challenges often grow with the volume of data, the number of systems, and the demands placed on how data is used. Legacy applications, differing business definitions, and data from multiple sources can make it difficult to establish a unified view.

AI makes data quality even more important. AI models work with the data sets they are given. If that data is inaccurate, incomplete, or inconsistent, the output cannot provide a reliable foundation of facts either. AI can accelerate analysis and support processes, but it does not solve fundamental problems in the input data.

For that reason, introducing AI should go hand in hand with the question of whether the underlying data is fit for its intended purpose. High-quality data, clear metadata, and a consistent data foundation help put results in context and enable their responsible use.

The Benefits of Good Data Quality: See More Clearly, Act With Confidence

The benefits of good data quality do not only emerge in large transformation programs. They begin in day-to-day operations: less manual reconciliation, faster analysis, fewer discussions about conflicting KPIs, and a better basis for decisions.
Improving data quality can help you:

• provide reliable KPIs across business functions,
• make data more efficient to use for analytics and reporting,
• identify errors and duplicates earlier,
• build automation on a more reliable foundation,
• better ensure that data is suitable for AI and analytics applications.

Not all data quality requirements carry the same weight. A shared monthly view may be entirely sufficient for certain management decisions. Other processes may require greater accuracy or more current data. What matters is that the requirements are transparent and that the quality of the data matches them.