Quality is More than Data

Fraud deserves the attention it gets. As the tools for generating fake or low-quality responses improve, so must our defences at recruitment and during fieldwork. That work is essential and must continue. But if we treat response authenticity as the whole of the quality problem, we miss the larger part of it, which now sits in how research is designed, how data is combined, and how findings are used.

Research Feeds the Wider System

Research has always informed decisions, but the detail, speed and reach of that influence has changed. Market research data feeds systems that set prices, personalise offers, decide who sees what, and shape the service a customer receives. Data gathered for one study is linked to other sources, used to train models or fed into automated decisions, often well after the project that produced it has closed.

The consequences are concrete. A segmentation can determine which offer a customer is shown. A pricing study can change how costs fall across different groups. A customer experience programme can decide who gets helped quickly and who waits. As datasets become more interconnected, it becomes harder to trace how any single study contributed to an outcome, and that is precisely when accountability tends to go missing.

Risk Accumulates Quietly

In this environment, risk seldom arrives as an obvious failure. It accumulates through a series of individually reasonable decisions. One team combines data from different sources. Another builds a model using data that was never designed for that purpose. Findings are reused well beyond the context in which they were collected, while partners assume someone else has addressed the gaps between them. Every step looks defensible on its own; the problem is the total picture.

This risk phenomenon stretches the researcher's role. The work now extends past analysis. We, researchers and insight professionals, are usually a step or even two removed from the final application, particularly on large or multi-partner projects. Yet the decisions we make about what to collect, how to frame the research, and how clearly we state the limitations still shape how the findings are interpreted and used. Responsibility travels with those decisions, however far we sit from the final outcome.

This matters more now than ever because the technologies that compound these effects are arriving quickly. In MRII's AI in Focus 2025 study, 62% of researchers reported that they or their team are now using AI, up from 40% a year earlier, and 77% held a favourable view of it. Adoption on that curve, combined with synthetic data, digital twins and automated decision making, means more research will be reused, modelled and acted on in ways the original designer never anticipated.

A Steady Principle, Actively Applied

We already have a strong foundation. The ICC/Esomar International Code sets out, as a fundamental principle, that all research must be conducted with due care and that interactions must be fair, respectful and avoid harm. That principle holds whether the method is a telephone survey or a study using synthetic data.

I say this from the inside. I was President of Esomar while the revised Code was being developed, so I share responsibility for what it says, and I now chair the committee responsible for how it is used. From that seat, the picture is clear, we have the principles and we need to apply them in deliberate ways that match our role.

The hard part has always been knowing what ‘due care’ means in practice. For example, when you are blending first-party data with a purchased model, utilising synthetic sample, or handing a dataset to a partner who will use it in ways you cannot see. That is why we are building an “ICC/Esomar Code Applied to ...” series. These will be short, practical interpretations that take existing principles and show how they apply to a specific, current problem. Synthetic data will be the first of these new formats. The principles endure as technology evolves, and the task is to show them at work in the situations researchers actually face today.

Due Care in Everyday Decisions

For most teams, the work is to make existing habits more deliberate:

• Ask how the data could be used beyond this project, and design with that in mind.
• Name the assumptions and gaps that could change the outcome and make them visible.
• Be explicit about the limits of the findings, including the uses they cannot support.
• Allocate responsibility across partners explicitly, so every gap has an owner.

A Broader Test of Quality

We have long defined quality as methodological rigour and accurate data, and that still holds. At Esomar we are working actively with the Global Data Quality initiative to raise the standard of the data itself, which matters a great deal. Quality, though, reaches further than data. It is also shaped by how research affects the people it represents and the decisions it drives. As research becomes more embedded in the systems around it, that dimension has become inseparable from the work. Defences against fraud keep poor data out. Due care, applied deliberately and demonstrated in practice, keeps us responsible for the quality of the data once it leaves our hands.

A Warning!

The scrutiny is coming whether we prepare for it or not. When a price, a segment, or an automated decision is challenged as unfair or biased, the underlying research gets examined, and the question is whether due care was applied and whether anyone can show it. A study built with that challenge in mind is more defensible and more useful to the client. One built without it becomes a liability waiting for a question.

So, build due care into quality itself, and design every study as though someone will later ask how the conclusion was reached, because increasingly they will. That habit protects your work, protects the client's decision, and protects the profession's licence to operate.