Data quality starts with people

‘What one person considers high‑quality data may be viewed as low‑quality data by others.’

The promise of data is clear: shared insights and better decision-making.
In practice, however, things are less straightforward. As soon as multiple departments start working with the same data, questions arise about quality, ownership, and interpretation. Not because the technology falls short, but because people think and act differently. Two experts share their perspective.

Maximising the value of data is seen in many organisations as a major promise and at the same time a recurring frustration. Translating theory into practice proves to be challenging.

Working within this tension are Marit Peters and Gert‑Jan Kooren from Capgemini Academy. Both work with data every day, but each from a different perspective. Together, they cover the key dimensions of the data challenge.

‘Data doesn’t solve problems by itself, but it does help you have better conversations and make better decisions.’ 

Marit Peters is a Learning Coordinator at Capgemini Academy. Her work focuses on training and adoption during large-scale organisational change. In her current assignment, she supports a nationwide organisation in its transition to a single SAP environment. She considers which target groups need to be trained, when, and how — ensuring that people can confidently work with the new system at go-live.

Gert-Jan Koren is a trainer and data expert at Capgemini Academy. Prior to this, he worked at Capgemini Insights & Data, where he was involved as an architect in designing data platforms and processes, and in setting up data governance and data quality management.

When is data considered high quality?

Gert-Jan: ‘Take an online webshop. As a customer, you create an account, place an order, pay, and receive the product. From an order-processing perspective, the data flow works perfectly if the order passes through the system without issues, the payment is processed, and the product is delivered.’ 

‘But the marketing department looks at this differently. They want to identify and track customers. How often does someone return? What do they buy? How does that behaviour develop over time?’ 

‘If we align data quality purely with the needs of the order department, there’s a good chance that when the same person registers again later using a different email address, two customer profiles are created. That’s no problem for order processing. But for marketing, it looks like two different customers — making customer insights far less reliable.’ 

‘What one person considers high-quality data can be low-quality data for someone else.’ 

‘I recognise that. People always enter data from their own work context.’ 

Could you provide an example? 

Marit: ‘If you ask people to fill in data fields without properly training them, they’ll enter whatever makes sense to them at that moment. That could be information useful to them — or simply something the system accepts, including answers like ‘not applicable’. Formally, the task is completed.’ 

‘It’s important to add: this is usually not unwillingness. People act logically within their own work. And there are personal differences too — some want to understand an IT system in depth, others want to spend as little time on it as possible.’ 

So data quality isn’t fixed. What does that mean for organisations?

Gert-Jan: ‘It’s essential that organisations clearly define what they need specific data for, and which quality requirements apply. Only then can you determine what is ‘good enough.’ 

‘That also means making choices. Not every dataset has to meet every possible requirement. Data can simply be good enough. Ask questions like: who uses this data, in which process, and for what purpose?’ 

Marit: ‘What I often see is that organisations leave these choices implicit. They assume people will understand — but that doesn’t happen automatically.’ 

‘If you don’t explain why certain data matters and what happens to it further down the chain, it remains an administrative task. Something people ‘have to do’, but not something that feels meaningful.’ 

Is governance the key?

Gert-Jan: ‘Governance is the starting point. From there, structure follows. Someone needs oversight and must consciously decide which data flows are leading — and which build upon them.’ 

‘In professional organisations, you often see a distinction between core data flows and derived flows. Core flows contain data that is used and trusted across the organisation — such as customer, product, or supplier data. These form the foundation for both processes and reporting and therefore must be consistent.’ 

‘Derived flows emerge when departments combine core data with additional information, add context, or create selections for their own purposes — for example, analyses, campaigns, or internal reports.’ 

‘You can compare it to a main road with side streets. The main road provides direction and coherence. Side streets are necessary to reach specific destinations, but they are not meant to reorganise traffic. They can be structured differently.’ 

‘Good enough data’ sounds pragmatic — but also difficult to manage. Why not aim for ‘as good as possible’?

Marit: ‘Let’s take an extreme but common situation. If you aim for perfection, small groups often emerge within organisations that focus entirely on data quality, convinced that if the data were perfect, the organisation would automatically function better — or new business models would suddenly appear.’ 

‘First, that’s rarely true: very few organisations have data at the core of their business model. And data can’t fix flaws elsewhere in processes. Worse still, striving for perfection leads to more data fields, more checks, more exceptions. Data management becomes complex, slow, and expensive.’ 

‘As a result, the original goal can be lost — or quality improvements in data cause quality loss elsewhere, for example because data becomes available much later.’

So how do you draw the line?

Gert-Jan: ‘It comes down to making choices. Decide: this specific piece of data must always be correct, because decisions depend on it. For other data, it may be sufficient that it provides direction, or that it can be enriched later.’ 

What behaviour is needed to embed data quality in practice?

Marit: ‘People need to be trained — and that requires a lot of clear communication. The key is helping people understand their role in the data chain. What happens to the data they enter? Who uses it later? Why does that matter? That’s how working with data becomes meaningful.’ 

How do you do this in practice?

Marit: ‘Giving meaning to the data journey is important for everyone involved. Then you look at what each target group needs. Not everyone needs the same knowledge. Someone entering data daily needs different insights than someone analysing it or making decisions based on it.’ 

‘In terms of training formats, many options are possible. Sometimes classroom training works well; often a short explanation at the moment of first use is far more effective. The goal is that people can — and want to — apply their knowledge when it becomes relevant.’ 

‘And after implementation, the work isn’t done. Only later do you see where people struggle, which shortcuts they develop — both positive and negative. That feedback is crucial for adjustment.’ 

Sounds simple…

Gert-Jan: ‘For me, empathy is the key word.’ 

‘An IT specialist, a marketer, a controller, and an executive all have different perspectives and interests. For each of them, high-quality data means something slightly different.’ 

‘As long as everyone reasons from their own role, people end up talking past each other. Only when you truly put yourself in someone else’s position and understand what they need from data can you make agreements that actually work in practice. That requires empathy. And you develop empathy—just as Marit mentioned—through communication.’ 

Finally: what is data quality, at its core? 

Gert-Jan: ‘Data is a means. A tool. Data doesn’t solve problems by itself — it helps you have better conversations and make better decisions, provided you’ve thought beforehand about how you intend to use it.’ 

‘Without those agreements, data remains an instrument that everyone expects something different from.’ 

Marit: ‘Most importantly, organisations must truly integrate data into everyday work — by continuously discussing why data is captured and how others use it.’ 

‘And by consciously keeping that conversation alive — between departments, roles, and people with different perspectives. Only then does data become something shared, rather than something owned by IT or an external consultant.’ 

Want to learn more about data quality in practice?