The GDQ benchmarking study found that when combining pre-survey and in-survey removals, research agencies are removing almost three in 10 ( 28.5%) potential survey responses, while suppliers are removing about one in five ( 21.2%).

The post-survey cleanout rate – the percentage of responses removed after survey completion – was found to be 8.9% for research agencies and 10.7% for suppliers.

Speaking at a GRBN/MRS webinar yesterday ( 23^rd June), Chuck Miller, Insights Association, presenting the results, said: “We are removing a pretty large proportion of sample. We need to look for better controls to improve quality, and the good news is that we’re scrutinising, and in the cases where we [the industry] feel we need to remove, we are removing.”

The research also found that B2B generally had a high removal rate – 22.5% combined pre- and in-survey removals, and post-survey removal was 59.6%. “This continues a trend that we saw in wave one – B2B audiences carry the biggest risk,” said Miller.  

The report benchmarks for the US, UK, Canada, Australia, Germany and France.

Buyer satisfaction with B2B sample has declined, compared with previous waves of the GRBN research, driven by a drop in the perceived quality of sample. The research found that for B2B, 43% of participating buyers were dissatisfied with online sample bought from third-party organisations in the six months prior to the study – an increase from 40% recorded in the previous wave, although the most recent wave had fewer participants (see methodology note). A fifth were neither dissatisfied or satisfied ( 21%) and 36% were satisfied.

Satisfaction with quality received a score of 78 – on a scale of 0-200, where 100 is average – for B2B sample (compared with 85 in the previous study).

Perceptions of quality were higher for B2C sample, at 91 – the same level previously recorded. For B2C sample, overall buyer satisfaction also remained the same, with 30% of participants dissatisfied (the same level as the previous wave), over half ( 55%) satisfied and 15% neither dissatisfied or satisfied.

The buyer sentiment survey also found that while just under a third of participants didn’t have an opinion on whether the use of AI impacted data quality in online sample, 34% stated that they felt AI decreases quality.

During a panel discussion on the results of the studies, GRBN’s Andrew Cannon discussed why buyers are increasingly questioning the value proposition of B2B sample. He said: “Either the price has gone up or the perception of the quality in relation to the price has decreased – my feeling is that it’s the latter, and the perception of the quality is being challenged.”

Brian Lamar, Insights Association Council for Data Integrity, said: “Absorbing operational costs is a factor that isn’t priced in. When routinely removing, the cost per complete is not really the true cost, which can be much, much higher, and I think buyers are doing the math with that.”

Anita Watkins, chief executive, IA, flagged the apparent contradiction between buyers still intending to buy online sample – this metric had not changed in the most recent research – while having a negative perception of quality.

Lamar said: “I think this a huge warning sign for our industry – I think buyers are probably looking at alternatives – either building their own panel, doing more qual or community work or looking at synthetic options. More credible alternatives are emerging, and that’s a concern of mine.”

Cannon added: “The industry probably has to up its game in terms of transparency and leading the discussion – not just letting the buyers lead the discussion around this.”

Debrah Harding, managing director, MRS, said: “It’s good there is still strong demand for research services but it would also be interesting to know who is buying the sample and to see the shift in buyers – in the UK, we are seeing increasing in-housing, DIY and so on. We potentially have a new group of buyers buying sample and some of those may be unfamiliar with the issues.”

For those who discover their benchmarks are below average, Harding advised making sure that their panel suppliers are supportive of the GDQ, and questioning whether they are committed to data quality and following best practice.

Harding also said it is important to consider client discussions. She said: “Are those questionnaires too long? Is the cognitive load being placed on participants too high? Are you boring your participants? It is not only about how you are performing overall, but looking at whether you have some clients whose projects are performing worse than others. Those client conversations need to be had.”

Methodology note

Global data quality benchmarking report (H1 2026 ): The Insights Association, on behalf of GDQ, collected data from October 2025-March 2026, with 46 participating companies across 13 countries providing random selections of pseudonymised survey data (around 1.8 million records).

GRBN online sample buyers sentiment survey: GRBN surveyed 151 buyers of online sample, with fieldwork in March–April 2026. The survey’s previous wave (conducted during the first half of 2025 ) had 346 participants.