Removal Rates and Recall Rates: The Two Numbers That Actually Tell You if a Panel Is Working
Most panel providers will tell you their size. Very few will tell you their removal rate. Here’s what those two numbers mean, what the GDQ benchmarks show us, and what to ask any provider before you place your next project.
Brian Lamar · Sr. Director of Sample Quality, ROI Rocket
There’s a question I wish more buyers would ask their sample providers, and almost none of them do.
Not “what’s your panel size?” Not “what’s your average CPI?” Not even “what fraud detection tools are you running?” — though that one’s getting closer.
The question is: what’s your removal rate?
And the follow-up: what’s your recall rate?
If your provider can answer both questions with a specific number — and explain what those numbers mean — you’re probably talking to someone who is actually managing a panel. If they can’t, or won’t, you’re probably talking to someone managing a mailing list and calling it a panel.
What these numbers mean and why they’re rarely published
A removal rate is exactly what it sounds like: the percentage of panelists removed from an active panel over a given period due to fraud, inactivity, failed quality checks, or identity verification failures. Our removal rate runs consistently between 5–10%. That’s not a problem to manage down — it’s evidence of active panel hygiene.
A panel that never shrinks isn’t being maintained. It’s being padded. When a provider quotes you a panel of two million respondents and can’t tell you how many of those have been active in the last six months — let alone how many have been removed for quality failures — that number is marketing, not methodology.
A recall rate measures whether the respondents you’re bringing back to a study are the same people who showed up before. In a tracking study, this is everything. You’re not measuring change in the market — you’re measuring change in the panel if your respondent pool is rotating without controls. A 90%+ recall rate target on tracking and longitudinal work is how you protect trend integrity.
Most providers don’t publish either of these numbers. Not because they don’t exist, but because publishing them invites the obvious follow-up questions: what’s your process? what triggers removal? how do you verify recall? Those questions separate providers who have invested in panel infrastructure from those who have built audience aggregators and layered fraud detection on top.
Panel size is marketing. Removal rate is methodology.
What the GDQ benchmarks tell us
That B2B post-survey cleanout number deserves its own conversation. A 59.6% post-survey cleanout rate means that for B2B studies capturing this kind of data, nearly six in ten completes are being removed after fielding. Whether that reflects the difficulty of B2B quality control, the concentrated fraud pressure in professional segments, or the specific standards of the providers contributing data to GDQ’s pool — it’s probably some combination of all three. But that’s a significant signal about how much invisible work is happening downstream of data collection itself.
GDQ’s May 2026 report, Safeguarding Qualitative Research, extends this analysis into qual territory — documenting how bad actors exploit online methodologies, often using AI to fabricate identities and responses, and how fraud pressure in qualitative recruiting has escalated alongside quantitative fraud. The report is a useful companion read to the benchmarking data.
The takeaway from GDQ’s benchmarks isn’t that these removal rates are alarming in isolation. It’s that removal is happening at scale in every study type, whether or not your provider is telling you about it. The question isn’t whether removals occur — they always do. The question is whether your provider is managing that process proactively and transparently, or quietly cleaning data after the fact and calling the output a deliverable.
Panel health is a qualitative story, not just a quantitative one
Here’s where removal rates and recall rates connect to something that matters beyond the numbers: what kind of respondent is actually participating in your research?
Panel health isn’t just a fraud management question. It’s a research quality question. And nowhere is that more visible than in qualitative work.
When you recruit for a focus group, an IDI, or a qual-forward study from a panel that lacks real identity verification and consistent removal protocols, you get respondents who are optimizing for incentives, not insight. They’ve learned what answers get them through screeners. They’ve done it dozens of times. They’re not lying, exactly — but they’re performing the role of research respondent, not living it.
A healthy panel produces qualitative respondents who are genuinely surprised when they’re selected, who haven’t been through your screener language four times before, and who engage with the research because they actually represent the population you’re trying to understand. That’s not a soft benefit — it’s the difference between research that reveals something and research that confirms what the incentive-optimized respondent thinks you want to hear.
GDQ’s Safeguarding Qualitative Research report makes this explicit: all segments of the research supply chain — clients, agencies, recruiters — share responsibility for mitigating fraud in qual, and the report calls for industry-wide collaboration to protect qualitative integrity. The fraud patterns described in that report aren’t hypothetical. They’re documented behaviors that active panel management — including regular removal — is specifically designed to prevent.
The recall rate matters differently for qual than quant. For a longitudinal qualitative study — re-engaging participants from a prior wave, running a brand-tracking study with open-ends, or revisiting a consumer community over time — knowing that the person coming back is actually the same person who showed up before is foundational. Without that, you’re not building insight over time. You’re building noise with a continuity illusion.
What to ask — and why it matters that you ask it
From our 10 Questions checklist — the ones that get directly at this
Our full checklist is a supplement to the ESOMAR 37, designed to reflect how the quality landscape has evolved since 2023. Three questions go directly to removal rates and recall:
Q1 What percentage of your active panel has taken a survey in the last six months? Panel recency is a proxy for panel health. A provider who can’t answer this precisely isn’t managing a panel — they’re managing a database.
Q4 What is your removal rate, and how often do you audit and clean your panel? A healthy panel has regular removals. If a provider’s panel never shrinks, it isn’t being maintained. Ask for a specific number. If they don’t track it, that’s an answer.
Q6 Have you measured how often your fraud detection system blocks legitimate respondents? This is the question most buyers don’t think to ask. Fraud detection makes two kinds of errors. A provider who understands both has thought carefully about their thresholds and why.
Download the full 10 Questions checklist →
A vendor who welcomes these questions is a different kind of partner than one who deflects them. That’s true in any vendor relationship, but it’s especially true in sample — where the quality of the underlying data asset is the entire foundation of the research you’re about to build on it.
The GDQ resources worth knowing
GDQ’s resource library at globaldataquality.org/GDQ-resources has become one of the most practical collections of data quality guidance in the industry. A few that are directly relevant to the removal and recall rate conversation:
Safeguarding Qualitative Research [May 2026] — GDQ’s most recent report on fraud in qual, including AI-fabricated identities
Data Quality Benchmarking Report — Wave 1 [Benchmarks] — the source for the removal rate data in this article
GDQ Buyers of Research Services Procurement Guide [Buyers] — practical quality signals for evaluating proposals and partners
Data Quality Transparency Checklist [Buyers] — project-level sample transparency questions to use before commissioning
GRBN Online Sample Buyers Sentiment Survey — Wave 1 [2025] — tracking how buyer sentiment on quality, price, and speed is evolving
GDQ Feedback Loop [Tool] — identify and classify quality issues in a consistent, transparent way
ESOMAR 37 Questions to Help Buyers of Online Sample [Foundation] — the foundational framework our 10 Questions supplements
The numbers that separate marketing from methodology
Panel size is marketing. Removal rate is methodology.
Fraud detection tool names are marketing. Recall rate on longitudinal work is methodology.
“Industry-standard quality controls” is marketing. A specific answer to “what percentage of your panel did you remove last quarter, and why” is methodology.
The market research industry has made real progress on the fraud detection conversation over the last few years. Buyers are more sophisticated. RFPs are more specific. GDQ’s benchmarking work has given the industry common language for discussing what good looks like.
But removal rates and recall rates are still the metrics most providers quietly avoid. They require transparency about panel maintenance that makes headline panel size numbers look smaller. They invite follow-up questions about process and infrastructure. They’re the difference between showing a buyer a clean-looking panel and actually running one.
We publish ours because we think that’s what being the data quality authority actually requires — not just claiming the title, but putting the numbers behind it.
If your current provider can’t or won’t do the same, that’s a useful piece of data in its own right.
Want the full checklist? Download 10 Questions to Ask Your Sample Provider Before You Place Your Next Project — a supplement to the ESOMAR 37 designed for how the quality landscape looks today. And explore the full GDQ resource library at globaldataquality.org/GDQ-resources.
