Data Quality in Market Research: Lessons from the UK Benchmark

In recent years, data quality in market research has shifted from a background concern to an industry-wide priority.

Reports of fake respondents, disengaged panellists and AI-generated survey entries have exposed just how fragile data integrity can be when quality checks fail.

In response, the Market Research Society (MRS), the Association for Qualitative Research (AQR), ESOMAR, and the Insights Association joined forces to launch the Global Data Quality Initiative (GDQ) – a coordinated effort to measure data quality and strengthen trust in research worldwide.

The initiative’s first audit, Wave 0, examined the quality of online quantitative surveys conducted in the United States and found that more than 40 percent of respondents still had to be removed for quality reasons during the survey itself, even after initial screening.

Building on that foundation, Wave 1 expanded the study internationally, covering markets including the UK, Canada, Australia, Japan and key European countries – to benchmark how data quality varies by region, methodology, and supplier type.

Wave 2, based on survey data collected between October 2025 and March 2026, gives us another look at the UK. Some findings reinforce the first report. Others need a more careful reading, particularly where the way data was recorded has changed. So, what do both waves tell researchers working in market research?

What the Numbers Actually Show

Wave 1 analysed over 280,000 UK survey records collected in the first half of 2025.

It compared data gathered directly by research agencies with data supplied through independent or third-party sample suppliers (organisations that provide respondent access for online or hybrid studies).

The UK Wave 1 report gave us a useful starting point:

UK findings from Wave 1

• Overall removals – Research agencies excluded 4.5% of respondents for quality or fraud concerns, compared with 11.1% among suppliers.* 
• Post-survey clean-outs – Around 5–6% of completed responses were removed for inattentive or low-quality behaviour.*
• B2C vs B2B – Consumer (B2C) samples showed removal rates well below global averages (6.6% vs 13.1%), whereas business (B2B) samples aligned more closely with international norms.*
• Secure survey links – Link encryption or another secure connection method was recorded for 99.2% of UK research agency records.*
• Abandonment – Only 6.7% of agency respondents dropped out mid-survey, compared with 14.9% for suppliers. 

*(Insights Association, UK Data Quality Benchmarking Report, Wave 1, 2025)

Those figures describe Wave 1. They should not be read as the latest UK benchmark.

What Wave 2 adds to the UK picture

The Wave 2 report adds five more UK findings:

• Overall removals – Suppliers stopped 18.6% of respondents before or during surveys, compared with 11.2% for research agencies. These were people removed by quality checks before completing the survey.*
• B2C vs B2B – Consumer (B2C) research had a combined removal rate of 8.4%, the lowest among the countries reported. The rate for business (B2B) research was 20.6%, although the B2B sample was much smaller.*
• Post-survey clean-outs – Agencies removed 6.5% of qualified completed responses for inattention or quality concerns, compared with 2.9% among suppliers. These rates are measured separately from removals before and during a survey.*
• A notable B2B result – 83% of UK B2B qualified completes were removed after the survey. The base for this figure was relatively small, so it should not be treated as a typical rejection rate for UK B2B research.*
• Secure survey links – Use of link encryption or another secure connection method was recorded for 56.9% of UK agency records and 25% of supplier records. However, this information was missing for 42.1% and 51.1% of those records respectively, limiting what we can conclude.*

It is tempting to line up every Wave 1 and Wave 2 percentage and declare that quality has improved or worsened. That would be misleading. Wave 2 clarified when removals are counted and reports some measures only for companies that recorded the relevant information. The two waves are useful together, but they are not a simple before-and-after test.

Beneath the Surface: Optimism or Oversight?

Lower removal rates can indicate genuinely better quality, but they can also signal less aggressive detection of poor behaviour.

Because every agency defines “fraudulent” or “low-quality” slightly differently, benchmarks depend heavily on internal processes and tools.

In both UK waves, suppliers reported higher combined pre- and in-survey removal rates than research agencies. That is a pattern worth examining, although the percentage alone cannot tell us whether one source has better participants or more stringent checks. 

Suppliers often aggregate respondents from numerous sources, and each hand-off introduces risk if vetting standards differ.

Three practical questions emerge:

1. Where do checks happen? Wave 2 separates removals before a survey starts, during the survey and after completion. That matters because two teams can report very different rates while checking for problems at different stages.
2. How visible is the recruitment chain? When a sample passes through several providers, researchers need to know where respondents came from and who checked their suitability.
3. What gets past automated checks? Even when overt fraud is removed, subtle disengagement – participants clicking through too quickly, providing shallow answers or failing attention checks – can still erode data quality. These behaviours often evade automated detection and require human oversight to catch.

At Angelfish, we apply multi-stage participant checks before a recruit reaches a session. The survey benchmark does not measure our qualitative recruitment, but its questions about sourcing, validation and accountability are familiar ones.

A Closer Look at Secure Survey Links

Wave 1 painted a strong picture of secure link use among UK research agency records. Wave 2 is less clear. It records use of link encryption or another secure connection method for 56.9% of UK agency records and 25% of supplier records. But this information was missing for 42.1% and 51.1% of those records respectively.

That means we cannot confidently say secure link use has fallen by a particular amount. We can say that the gaps in reporting make it harder to assess a basic quality control. For researchers commissioning online surveys, it is worth asking how survey access is protected and how consistently those checks are recorded.

How Brands Are Responding

Across the industry, concern has evolved into expectation.

In Research Live’s feature “What Do Brands Think About Data Quality?” (2025), client-side researchers reported that they now expect proof of quality, not just promises.

An open letter published by MRS and industry leaders echoed that call, urging agencies and suppliers to collaborate rather than compete on quality standards (Research Live, 2025).

For brands, data quality is no longer a differentiator; it is the minimum requirement. For agencies, credibility now depends on being transparent about how integrity is achieved from recruitment to analysis.

The benchmark gives clients more useful questions to ask: Where did participants come from? What checks happened before and during fieldwork? What was removed afterwards, and why?

The Qualitative Challenge: Where “Good Enough” Isn’t

The GDQ benchmark draws on online survey projects. It does not measure the quality of qualitative interviews, focus groups or research communities. Still, the questions it raises have clear implications for qualitative fieldwork.

Authenticity – the tone, emotion and depth that define qualitative insight – cannot be guaranteed by automation alone.

Four pressures continue to shape data quality in qualitative fieldwork:

1. Price pressure and procurement-first thinking – When recruitment is treated as a commodity, timelines and budgets squeeze validation efforts.
2. Over-reliance on automation – Fraud-detection software helps, but AI-generated or semi-automated responses are becoming increasingly sophisticated.
3. Participant fatigue – Over-surveyed audiences, low incentives and impersonal onboarding reduce engagement before fieldwork even begins.
4. Opaque recruitment chains – Multi-tier supplier models can make it difficult to trace where a participant originated or how they were screened (MRS Data Integrity Report, 2024).

At Angelfish Fieldwork, we see these factors every day, and know that the solution lies in human-centred recruitment.

Personal communication, active validation and transparent sourcing create the trust that automated systems alone can’t replicate.

Why the UK Benchmark Matters

The value of the UK figures is that they make quality questions more specific. Wave 2 shows differences between agency and supplier records, and between consumer and business studies. It also shows how much a missing measure can limit what we can conclude.

But benchmarks are not just a scorecard, they are a call to action.

A lower removal rate is not automatically a mark of better quality. It might reflect stronger recruitment at the start, or it might mean fewer problems were detected. To understand which, researchers need to look at the checks behind the number. That could include:

• Monitoring response speed and consistency during screeners or interviews.
• Analysing open-ended language for signs of repetition or non-human phrasing.
• Combining digital checks with manual review where anomalies appear.
• Asking suppliers to explain where respondents came from, which checks were applied and when removals happened.

Together, the two waves give UK researchers a way to have a more informed conversation about quality, backed by evidence rather than assumption.

Angelfish Fieldwork’s Perspective: Beyond Compliance

At Angelfish Fieldwork, quality begins with people, not platforms.
We follow a simple principle: every great project starts with the right participants.

We:

• Verify each participant’s identity and suitability using a blend of manual and digital checks before fieldwork begins.
• Maintain transparency in every recruitment chain so clients always know the source of their respondents.
• Engage participants personally to confirm motivation and relevance, ensuring genuine interest in the topic.
• Gather feedback after each project to refine and improve our recruitment approach.

In practice, that means data integrity isn’t an isolated checkpoint, it’s woven through every stage of recruitment.

As Lisa Boughton, Co-Founder and Director of Angelfish Fieldwork, said when we first explored the Wave 1 findings:“Wave 1 gave us the benchmark; now it’s up to agencies like ours to raise it.”

What the Industry Should Take Away

Both waves focus on online quantitative research, but their lessons are useful for anyone responsible for finding, checking or working with research participants.

Four priorities emerge for the UK research community:

1. Recognise that quality is structural. It must be built into recruitment and design, not audited afterwards.
2. Strengthen governance. MRS, AQR and ESOMAR are driving frameworks, but responsibility also sits with every agency commissioning or supplying respondents.
3. Prioritise participant experience. Engaged participants deliver richer, more accurate insights.
4. Collaborate on standards. Sharing best practice openly benefits the entire research ecosystem.

What Happens Next?

Wave 2 gives us another UK benchmark, but it does not settle every question. The figures show where to look more closely; understanding why a response was removed, or why a measure was not recorded, still takes openness from the people collecting and supplying the data.

For researchers, that means asking for a clear account of participant sourcing and quality checks, and making space for those checks in the brief, timeline and budget. For qualitative recruitment, it also means speaking to people and checking their suitability before they take part, rather than assuming a completed screener tells the whole story.

Because if we can’t trust the people behind our data, we can’t trust the strategies built on it.

Want to learn more about data quality in market research?

Explore our Market Research Data Quality page or revisit our earlier blog, How Fieldwork Agencies Can Strengthen Data Quality.

Looking for a partner that takes respondent integrity seriously?

Discover how we recruit authentically on our Participant Recruitment page.