The state of online survey data quality
The Global Data Quality (GDQ) Data Quality Benchmarking Project is a global research-on-research initiative to understand and improve data quality. Wave 2 covers data collected October 2025 through March 2026, contributed as random, pseudonymized selections of online survey data by participating companies across the research, insights and analytics industry.
45
Companies
13
Countries
~1.8M
Records
6
Detailed markets
4
Study types
Global headline: combined pre-survey and in-survey removal
Weighted. Pre-survey plus in-survey (total) removal, computed before rounding.
Key findings
Wave 2, global
Where to go next
This report is built to be cut, not read front to back.
Reading the numbers
Every figure in this report carries its base size. Flags are applied automatically, not editorially.
LOW Low base 500 to 4,999 records — directional only
V-LOW Very low base under 500 records, or 75%+ missing — not reportable
N/B Base not published figure carried from the summary tables
— not reported in this wave
Hover or tap any flag to see the base behind that figure. Many Wave 2 metrics are reported only among companies that record the relevant data, so bases differ metric to metric within the same cut. Link encryption is the least complete metric in this wave, with 23% to 95% of records missing the field depending on the cut.
Explore Benchmarks
Pick a market and a cut. Every view below redraws, including the findings.
Quality Removal Benchmarks
Combined pre-survey and in-survey (total) removal
Headline tiles = pre-survey + in-survey (total) removal, computed before rounding; may differ from the sum of displayed stages by 0.1 point.
Quality Benchmarks
Incidence, LOI, abandon rate, device, and encryption
Country Comparison
All 13 participating markets on one screen. Sort any column; expand a row for the full profile where detailed data exists.
Market comparison
Shading runs light to dark within each column. For removal and cleanout, darker means more removal. For link encryption, darker means higher adoption.
Study Types
The cut that moves most. B2B carries the highest quality risk in every wave so far.
Global combined removal by study type
Unweighted. Pre-survey plus in-survey (total) removal.
Full global metric set by study type
Unweighted
Study types by company type
Mix of contributed records by self-defined company type
B2C continues to dominate the record mix for both company types. Suppliers contribute proportionally
more B2B records (10.9% vs 7.2% for Research Agencies), while Research Agencies edge ahead on Healthcare Provider work.
Market coverage by study type
Which markets reached reportable sample in Wave 2
B2B coverage is markedly narrower than B2C. Healthcare study types reached reportable
sample only at the global level and, for Healthcare Patient, in the US.
Benchmark Yourself
Enter your own rates to see where you sit against the benchmark. Nothing you type leaves this page; there is no server behind this report.
Your rates
Enter percentages. Leave any field blank to skip it.
Where you land
Benchmark, don't judge. A rate above or below the benchmark is a prompt to look at your process,
not a score. Removal rates reflect how much bad traffic you catch and how much reaches you in the first place.
Methodology & Appendix
How Wave 2 was built, what each metric means, and what changes in Wave 3.
About the GDQ Data Quality Benchmarking Project
The Global Data Quality (GDQ) Data Quality Benchmarking Project is a global research-on-research initiative to understand and improve data quality. It seeks to define and track a set of standards or norms to be used to assess data quality at both the industry and individual company levels. With these norms, this initiative aims to contribute to the development of globally accepted data quality buying signals and fuel innovation that strengthens the research ecosystem.
How it works
This initiative can only be done through industry collaboration, with the participating companies in the research, insights and analytics industry (including brands, research agencies, panel providers, tech platforms, digital analytics and consultancies) contributing random sets of data from online survey projects. Metrics are aggregated in various ways to assess the state of research data quality.
Methodology
Data collected October 2025 through March 2026
- Participating companies provided random selections of pseudonymized survey data.
- Wave 2 retains global scope and broadens country-level reporting to include the US, UK, Canada, Australia, Germany, and France.
- Weighted data are used for Research Agency vs Supplier views; unweighted data are used for study-type views.
Company type Research Agency vs Supplier (weighted)
Study type B2C, B2B, Healthcare Patient, Healthcare Provider (unweighted)
Markets Global plus US, UK, Canada, Australia, Germany, France in detail
Many Wave 2 metrics are reported only among companies that record the relevant data
(for example pre-survey blocks, in-survey fraud distinction). See the data mapping below.
Progress
Wave 1 to Wave 2
Submission metadata
Participating companies and reporting coverage
Who administers pre-survey controls (among Research Agencies)
Key findings
Sample Suppliers report near-universal pre-survey controls and are most likely to record removals and distinguish fraud. Among Research Agencies, just over half rely on their Sample Supplier to administer pre-survey controls, while roughly a third use both their own controls and the supplier's.
The weighting appendix reports Research Agencies n=24 and Suppliers n=21, totalling 45 companies.
The submission-metadata breakout above groups Platform Companies separately from Sample Suppliers; the two views
use different supplier groupings.
Data mapping for benchmarks
Appendix
Subgroups: Research Agency = self-defined Research Agency. Supplier = self-defined Sample Provider
Company, Quality Tech Company, Platform Company, or Other.
Approach to weighting
Appendix
- Wave 2 retains the 2-cell, 3-tier weighting structure introduced in Wave 1 to ensure the sample is balanced and can be assessed consistently across waves.
- The weighting mechanism is based on company size, using the self-reported number of completes as the indicator. This ensures larger companies have appropriate representation while preventing smaller companies from being weighted too far down.
- Sample in many countries remains small. Cross-region and in-region representation will continue to be addressed in future waves.
Weighted data are used for total and country-level Research Agency vs Supplier analysis.
Unweighted data are used for study-type and subgroup analyses.
How to cite this report
Every figure in this report belongs to a cut and a base. Quote all three and the number
travels safely.
The report
A single figure
Before you quote a number
- Name the cut. A removal rate means nothing without knowing whether it is Research Agency or Supplier, which market, and which study type. "The industry removal rate" is not a figure this report produces.
- Carry the base. Every figure here shows the number of records behind it. Quote it.
- Respect the flags. A figure marked LOW is directional and should be quoted only with its base stated. A figure marked V-LOW rests on fewer than 500 records, or a field missing from three quarters of records, and should not be quoted as a benchmark at all.
- Do not compare straight to Wave 1 on removal metrics. Wave 2 revised the definitions: in-survey fraud and behavior are now reported together, and removal is reported among companies recording each stage. Those changes are not like-for-like against Wave 1.
- Link rather than retype. The address bar carries the view you are on, so you can send a colleague a link to the exact cut instead of a screenshot of it.
Questions about a figure, or a use that is not covered here, go to the GDQ Benchmarking
working group before publication.
How to use the data
- Benchmark, don't judgeUse these metrics to understand how your data compares with industry norms, not as an absolute measure of success or failure.
- Segment comparisonsCompare your study outcomes against relevant benchmarks (for example by supplier type, research agency, country, or study type).
- Identify gapsLook for areas where your results diverge significantly from the benchmarks; these may indicate opportunities for process improvements or quality innovations.
- Track progressApply these benchmarks across time to measure your organization's progress wave to wave.
- Collaborate and contributeYour participation not only helps your organization; it refines future benchmarks and ensures the data reflects the broadest possible view of industry standards.
Next steps
Wave 3: what's changing
- Data submissions move to a 6-month cadence with fixed annual deadlines: October 15th and April 15th. Mark your calendars.
- Under consideration: collecting aggregated survey-level metrics instead of respondent-level records. This widens who can pull and deliver data, opens participation to brand-side researchers, and removes anonymization concerns around individual-level records.
- Share feedback with the working group ahead of the Wave 3 call for data.
Wave 2: benchmark focus areas
- Reducing reliance on post-survey removal, especially in B2B where rates remain elevated.
- Improving completeness of link encryption reporting (roughly half of records are currently missing this field).
- Investigating country-level anomalies where actual IR exceeds sold IR (Germany Research Agency, Canada B2C).
Why participate
- Gain a quantitative understanding of what "good" data quality looks like.
- Compare your studies' performance against the industry overall, by country, company type, and study type.
- Identify white spaces against which your company might innovate.