Data quality at every point a session can go wrong
No single check catches everything. ResearchDart layers controls before, during and after the survey, and records a reason for every decision so buyers and suppliers can see where quality was lost.
Every control, and what it records
Each row is a decision the router can make about a session, and the reason it writes down when it does.
| Stage | Control | What it stops | Recorded as |
|---|---|---|---|
| Entry | Signed entry links | Respondent IDs edited in the address bar | Refused, nothing recorded |
| Entry | Respondent ID deduplication | The same supplier ID entering a project twice | duplicate_respondent, GDQ 4 |
| Entry | Device deduplication | One browser entering under several IDs | duplicate_device, GDQ 4 |
| Entry | Automation checks | Headless and scripted browsers | automation, GDQ 1 |
| Screening | Provider session scoring | AI agents, bots, VPN and datacenter traffic, multi-accounting | automation or anonymized_network or fraud_tool |
| Screening | True location | Connections outside the project country | geo_mismatch, GDQ 3 |
| Survey end | Signed endlinks or status API | Forged or edited complete URLs | Refused, nothing recorded |
| Survey end | Minimum LOI | Completes faster than a person can read the questions | speeder, GDQ 6 |
| Survey end | First outcome wins | Replayed endlinks changing a recorded outcome | Original outcome kept |
| After field | GDQ-coded reversals | Completes that fail cleaning | Any of the 18 GDQ codes |
Recommended starting settings
Adjust after soft launch, once you have real timings and screening results.
- Screening mode
- Monitor for the first project with a new supplier, then enforce.
- Suspicious sessions
- Allow and flag for consumer studies; block for B2B and other high-incentive studies, which show the highest removal rates in the 2025 industry benchmark.
- Minimum LOI
- About one third of the median LOI from your first 50 completes.
- Duplicate devices
- On, except for studies in shared locations.
- Signed endlinks
- On whenever your survey platform can sign server-side; otherwise use the status API.
Data quality questions
What is a quality terminate?
Does ResearchDart replace in-survey quality checks?
How is a duplicate device identified?
Why can some duplicates be legitimate?
Further reading
A survey data quality checklist for 2026: before, during and after fieldwork
The controls worth having in an online survey this year, sorted by where they run. Short on theory, specific about settings, and honest about which ch...
Data QualityAI agents now pass attention checks. Plan your fieldwork accordingly
A research team built an AI respondent that passed 99.8% of attention checks. That does not mean online sample is finished. It means the checks most s...
Research OperationsWhat the data quality benchmark says about your feasibility estimates
The Insights Association and GDQ benchmark pooled close to two million survey records from 51 companies. Most coverage quoted the removal rates. The m...
Put a quality layer in front of your next study
Tell us whether you buy sample or supply it. We set up your account, share sandbox links and walk through the first integration with your team.