Integrity Screening
Check 100% of your interviews, from the first day of fieldwork.
Every recording is checked against your questionnaire and answer data, and the whole project for interviewer patterns. Falsified, cut-short and incomplete interviews are flagged within minutes; your QC team listens at the exact moment and makes the final call.
Questions not asked, answers recorded
Interview 40276 · Interviewer #22 · Day 2 of field
- 05:35–06:20
Q6 · Local services ratings
List not read, all three ratings recorded
- 08:02–08:30
Q11 · Household size
Not asked, answer recorded
- 09:12–09:40
Q14 · Age group
Not asked, answer recorded
Your reviewer listens, then decides
What gets an interview flagged
No interview
Answers recorded, but no respondent answering on the recording.
Cut short
The respondent is gone, and the answers keep being recorded.
Questions skipped
Answers recorded for questions that were never asked.
Answers that don’t match
Recorded answers that differ from what the respondent said.
What we check
See what happened inside the calls, not just how long they took.
Two analyses feed one list. Call analysis listens to each recording; pattern analysis reads your answer data across the project. Together they show which interviewers skip questions, record answers nobody gave, or run interviews impossibly fast.
Call analysis
Every recording, against your questionnaire:
- Was there an interview: a respondent who answers, and time enough for the questions?
- Was it finished, or cut short with the remaining answers recorded anyway?
- Were the questions asked, including the ones the answers route to?
- Do the recorded answers match what the respondent said?
Pattern analysis
Your whole answer export, across interviews:
- Interviews by one interviewer that overlap in time
- Interviews shorter than it takes to read their questions
- Answers copied from another interview
- An interviewer with far more of the fastest interviews than chance allows
| Interviewer | Checked | Flagged |
|---|---|---|
| #08 | 46 | 1 |
| #14 | 39 | 2 |
| #22 | 12 | 9 |
| #31 | 41 | 1 |
Why #22 stands out
- 6 interviews questions not asked, answers recordedCall analysis
- 2 interviews no respondent heardCall analysis
- 7 of 12 among the project’s fastest 5%Pattern analysis
Length alone would only say #22 is fast. The calls say why the interviews need a listen.
Evidence, not scores
Your QC team gets a to-do list, not a sample.
Instead of picking interviews at random, your reviewers start with the ones that need them, and know exactly what to check in each. A flag means “listen here”, never a verdict: your team confirms or dismisses it.
Needs review
6 of 214 interviews checked today
- No interview
Only the interviewer is heard, yet 41 answers are recorded
Interview 40213 · Interviewer #22 · Call analysis
Listen at 00:00–02:10 - Critical
Respondent hangs up at 3:12; your data has the interview running 11 minutes
Interview 40355 · Interviewer #14 · Call analysis
Listen at 02:50–03:12 - Critical
Q6, Q11 and Q14 not asked, answers recorded
Interview 40276 · Interviewer #22 · Call analysis
Listen at 05:35 +2 more - Critical
4 recorded answers differ from what the respondent said (Q15, Q16, Q21, Q22)
Interview 40302 · Interviewer #09 · Call analysis
Listen at 07:48 +3 more - Critical
Overlaps the same interviewer’s interview 40322 by 6 minutes
Interview 40318 · Interviewer #31 · Pattern analysis
From your answer data - Needs review
7 of this interviewer’s 12 interviews are among the project’s fastest 5%
Interview 40331 · Interviewer #22 · Pattern analysis
From your answer data
The exact minute
Each flag names the interview, the questions and the times to listen to in your own recording. No scrubbing through a 15-minute call.
Most serious first
Interviews with no real interview behind them top the list, then interviews cut short or with questions skipped.
Short by design
Only gross problems flag an interview. A single slip stays in the record, so every flag on the list is worth a listen.
From day one
Catch bad completes before your client does.
Checking starts the day your project goes into field. A problem a sample takes weeks to find, if it finds it at all, shows up in the first days, while you can still retrain, remove and replace.
Checked as they come in
Send recordings and answer data as interviews finish, or have them pulled automatically.
Replace inside the field period
Confirmed bad completes are re-fielded while quotas are open and the deadline is still weeks away.
Nothing passed by default
An interview with no recording, or one split across files, is reported as not checked, never counted as clean.
A record for every complete
Export what was checked, what was flagged and what your team confirmed, ready to share when your client asks how the data was verified.
Day 1 · 09:00
Fieldwork opens
Day 1 · 22:00
38 interviews checked, 2 need review
Day 2
Interviewer #22: 4 of 6 interviews flagged
Day 3
#22 retrained; 5 confirmed interviews re-fielded
Day 12
Field closes: 1,174 of 1,212 completes checked
How it works
Set up in days. Checking from the first interview.
- 01
We set up your survey
Send the questionnaire and a sample of your answer export. We map the questions, routing and answer codes, and you approve the result. Days, not an integration project.
- 02
Upload, or connect once
Upload recordings and answer data in batches, or connect your CATI or recording system once and interviews are pulled automatically as they finish.
- 03
Every interview checked in minutes
Each recording is checked against your questionnaire and your answer data, and the whole project for interviewer patterns.
- 04
Your team listens and decides
Reviewers listen at the times given and confirm or dismiss each flag. Nothing is final without a human decision.
- 05
Replace before delivery
Confirmed bad completes are re-fielded while the field is open, and you deliver with a record of what was checked.
Security & data
Your clients’ data, defensible.
Research agencies process their clients’ data, so your vendors have to survive your clients’ due diligence. VeriCATI is built for that, whichever product you run.
- No training on your data
- Nothing you upload trains public or shared models. Enterprise agreements with our AI providers bar training on your data and require deletion from their systems, typically within 30 days.
- Recordings kept only as long as needed
- Integrity Screening keeps no recordings once an interview is checked. Deep Quality Control can delete raw recordings a set number of days after processing, keeping only the findings and reports.
- Per-account isolation
- Your data is scoped to your account; no cross-customer access.
- NDA and DPA before any data moves
- Signed before we process your first batch, including demos and pilots. The list of AI providers that process audio and transcripts is shared on request.
FAQ
Questions, answered.
Something else on your mind? Send us a question below.
Does a flag mean the interviewer falsified the interview?
How soon is an interview checked?
What do you need from us?
What about interviews without a recording?
Does it replace live monitoring?
Do you keep our recordings?
Can we try it on a finished project first?
How is it priced?
Try it on a project you’ve already finished.
Send one completed project: the questionnaire, the answer data and the recordings. We’ll check every interview and show you what the sample didn’t.