Data quality is an end-to-end discipline built on four pillars: Preventing design flaws, Protecting against fraudulent entry, Purifying data with internal signal vectors, and Proving quality through transparent, actionable metrics—all reinforced by human-in-the-loop oversight.
The Data Quality Report brings that 4P framework into one place, providing the evidence for every study so you can see exactly what was screened, quarantined, and verified before your results reached you.
⚠️ Note: The Data Quality Report is enabled per account. If you don't see the Quality tab on a panel survey, ask your aytm contact about turning it on.
In This Article
- Terminology
- Finding the Report
- The Five Sections
- How Responses Are Removed
- Reading the Numbers
- Troubleshooting
Terminology
| Term | Definition |
|---|---|
| Quality | The tab in your results sidebar. Opening it reveals the Data Quality Report. |
| Data Quality Report | The full panel behind the Quality tab—five sections covering the whole quality process. |
| Data Centrifuge | One section inside the report—the automated, response-level analysis engine. Not the tab, and not the whole report. |
| Quarantine | A response the Centrifuge held out of your dataset automatically. Provisional and reversible. |
| Removal | A response taken out by a person—either you or an aytm expert. A settled decision. |
| Blocked attempt | An entry attempt stopped before the survey began. It never became a response. |
Finding the Report
From your survey's results page, click Quality in the left sidebar.
The report loads with the PaidViewpoint proprietary panel section already expanded and the rest closed. Sections open and close independently, so you can have any number of them open at once. Your open sections stay put if you switch to another stats tab and come back, and reset when you reload the page.
⚠️ Note: Quality is the tab, Data Quality Report is the panel it opens, and Data Centrifuge is one section within that panel.
Which surveys have it
The report describes a panel-sourced respondent population, so it only appears on panel surveys. Three situations mean no Quality tab is created at all—you won't see a greyed-out tab or an upgrade prompt, just no tab:
- The survey wasn't fielded to the aytm panel—for example, a survey sent to your own list.
- The survey is still in draft. The tab appears once you launch.
- The survey was assembled from other surveys' responses, such as a merge or a clone taken together with its source's responses. A clone taken without responses is an ordinary survey and gets the full report.
If you're viewing a survey through a share link, you'll see the report at the level the survey's owner set, and the export option won't be available to you.
The Five Sections
The report always shows the same five sections in the same order. Each row carries a one-line summary with that section's headline figure, so you can read the whole picture without opening anything.
1. PaidViewpoint proprietary panel
Quality starts with the panel itself. This section covers PaidViewpoint's track record, links to SurveyPolice and the panel itself, and explains blended mode—where PaidViewpoint is augmented with Trusted Partner Panels to extend reach across 100MM+ consumers.
This section reads nothing from your survey. It's the same on every survey, and it doesn't indicate which panels sourced your responses—aytm doesn't disclose a survey's panel composition. Think of it as background on the standard your respondents come from.
2. Traffic Sentinel
Traffic Sentinel is the first line of defense, controlling inbound traffic and preventing fraudulent participants from reaching your survey. It combines device fingerprinting, SuperCookies, graph theory, and network intelligence to identify users reliably—tracking devices even in private mode, maintaining persistent IDs across sessions, and detecting VPNs and proxies, while staying privacy-compliant.
Everything counted here happened before the survey. These are attempts that never became responses, which is why they don't appear anywhere in your results.
3. Data Centrifuge
The Centrifuge is deep, response-level analysis of the responses you did receive. It groups its findings into categories—checks such as twinned responses, pacing anomalies, response patterns, attention checks, and verbatim anomalies—and shows how many respondents each one affected.
It runs as your survey fields, so figures accumulate live and the summary line tells you when they're still provisional. Once fielding closes, the wording changes to confirm the analysis covered your entire sample.
Not every category fires on every survey. A check with no input stays silent—no open-ended questions means no verbatim checks, and no conjoint means no model-fit checks.
4. Human review
This section covers the removals a person made. It always shows two things: the responses you rejected on open-ended answers, and the panelists you gave Kudos to.
Rejecting a response limits its author's ability to participate in future surveys. Giving Kudos does the opposite—it motivates respondents to keep leaving meaningful answers, and aytm pays a surprise bonus to the most useful respondents as a gesture of appreciation.
A third block appears when your survey received an additional expert review. In that case, trained researchers reviewed open-ended answers to detect inattentive or fraudulent patterns, and any responses they judged inaccurate are reported here.
5. Survey performance and feedback
This section leads with people rather than rates. Drop-offs and exits tells you how many respondents entered your survey and didn't finish, with a chart showing where they left. Each point on the chart is a number of respondents at a specific question, and only questions that lost at least one respondent appear—so a long survey doesn't fill the chart with empty points. Hover a point to preview that question's wording.
You're never charged for drop-offs. And because pre-qualification outcomes are a different thing entirely, they live in the Term Report—there's a link beneath the chart that takes you straight there.
Positive Feedback Rate is the average of the 1–5 star ratings respondents leave after completing your survey, shown next to the platform average so you have something to compare against. Treat that comparison gently: respondent feedback can be affected by factors outside your control. The average is withheld until you have at least five ratings, so a handful of responses can't produce a misleading number.
How Responses Are Removed
The report deliberately keeps three removal paths separate, and they're reported under different names in different places. Most "these numbers don't add up" questions come from adding together figures that were never meant to be summed.
| Path | Who acts | Where you see it |
|---|---|---|
| Automated quarantine | The Data Centrifuge, with no human involved | The Data Centrifuge section's count, and the quarantined-responses export. Not in Human review. |
| Expert removal | aytm researchers | The expert-review block in Human review, and the header's Human Review removals. |
| Your own rejection | You, on open-ended answers | The "You've rejected" count in Human review, and the header's Human Review removals. |
The difference that matters: a quarantine is provisional and reversible, while a removal is a settled decision. That's why the two are counted under separate names on the same page instead of being added together.
The counts at the top of the page
Whenever the Quality tab is open, two extra counts appear beside your survey's N and completion status. They're a running tally, so you can see the headline numbers without opening a section.
Human Review removals combines the two people-driven paths above—aytm expert removals plus your own rejections—into a single number. Automated quarantines aren't in it; those stay in the Data Centrifuge section.
Blocked survey attempts counts entry attempts that Traffic Sentinel stopped before the survey began, so nothing was taken out of your data—those people never became respondents in the first place. It's a count only, with no percentage beside it, and a screening pass that blocked nobody reads 0.
Either way, you never pay for them. aytm's billing promise covers both the responses that were quarantined automatically and the ones removed by a person.
Reading the Numbers
A few conventions run through the whole report, and knowing them saves a lot of second-guessing.
An em dash (—) indicates no figure yet, not that the figure is zero (the zero itself can be a figure). You'll see it where a platform average hasn't accumulated enough surveys to be meaningful.
- A zero is a real measurement. "0 blocked attempts" is good news.
- Rates at 1% and above show as whole percentages (17%). Below 1%, you'll see one decimal (0.4%), and below 0.05% you'll see an upper bound (<0.1%) so a real, non-zero figure is never captioned as zero.
- Where a figure doesn't apply, it's dropped rather than shown empty. If no completion rate is available, the drop-off line states the count on its own with no percentage in brackets.
What the drop-off share measures
The drop-off share is the exact complement of your completion rate. Both use the same denominator—everyone who entered the survey body, including those who finished over quota—so the share and the count beside it always describe the same group of people.
Screening outcomes aren't in that group. Pre-qualification screen-outs, trait screening, and country-selector walk-aways are all screening results, and they surface under Traffic Sentinel and in the Term Report instead.
⚠️ Note: Filters and segments don't affect this report. Applying, changing, or clearing a filter on your charts leaves every figure on the Quality tab unchanged—the report always reflects your survey's whole response set.
Troubleshooting
While your survey is still fielding
The report is available throughout fielding, and the Data Centrifuge updates as responses come in. Its summary line says so directly, and its figures can change until fielding completes. A survey closed by refund counts as complete, the same as one that finished normally.
The report won't load
You'll see Loading... in place of the summary lines while the report fetches. If it fails, the section header offers a retry that reloads the whole report. If a refresh fails after the report has already loaded, your existing figures stay on screen rather than being replaced with an error.
The drop-off chart is empty
Drop-off tracking is a new capability, available for surveys launched on or after August 13, 2026. Surveys launched before that date may show limited or no data here.
This isn't a delay—the data isn't backfilled, so a survey that closed before that date will always show this message. You'll also see it, correctly, on a survey where nobody abandoned partway through.