Skip to content

Latest US poll: 63% say gas prices are affecting their household budget a great deal or a moderate amount

See the poll

Data quality · FraudCheck

FraudCheck: stop fake respondents at the door

FraudCheck runs as each person opens a CatalyseAI study, before they answer a single question. It stops repeat entries, people outside your country and traffic from VPNs or proxies, and you decide per survey whether a failed respondent is terminated or only flagged.

  • Runs before question one
  • Rules for each survey
  • Terminate or flag, with a reason
6checks you switch on for each survey
0–100threat score: 0 is clean, 100 is fraud
24hthe default window of the duplicate-IP check
2actions for a failed check: terminate or flag

Why it matters

Bad respondents are cheapest to stop at the start

Some people take the same survey twice from a second browser. Some sit behind a VPN or a data-center address. Some are in a different country from the one you are paying for. If they get in, you pay for their answers, spend time finding them, and then fill their quota places again. FraudCheck moves that work to the moment they arrive.

The door

Six checks between a shared link and your data

Follow 1,000 people opening a survey link shared on social media, one check at a time.

  1. Duplicate IP
  2. Duplicate device
  3. Country
  4. VPN and proxy
  5. Threat score
  6. Length limits
Step 01 / 06

Duplicate IP

Rejects a second start from an IP address that already started or completed the survey within a window you set, 24 hours by default.

203.0.113.24 · started 09:02In
203.0.113.24 · starts again 14:40Same IP within 24 hours
979still in21 second starts from one IP address
Step 02 / 06

Duplicate device

Rejects a browser that has already taken this survey.

Browser fingerprint7f3a 91c2 e04b
Seen before
Cookies clearedNew IP addressStill the same browser
965still in14 browsers already seen
Step 03 / 06

Country mismatch

Rejects people whose IP address is outside the countries you allow, for example US only.

Allowed: United StatesYour setting
198.51.100.43 · GermanyOutside the allowed countries
932still in33 outside the United States
Step 04 / 06

VPN and proxy

Rejects VPN, proxy, Tor and hosting-provider addresses.

VPNProxyTorHosting providerHome or mobile connection

The IP address is looked up as each person arrives.

913still in19 VPN, proxy, Tor or hosting addresses
Step 05 / 06

Threat score

Rejects addresses whose risk score, from 0 (clean) to 100 (fraud), is at or above the limit you set.

Your limit: 7512 in88 stopped
0 clean100 fraud
907still in6 at or above your limit
Step 06 / 06

Interview length limits

Treats completes faster than your minimum as speeders and slower than your maximum as distracted. Leave either empty for no limit.

Min 6Max 453: speeder18: in58: distracted
0 min60 min
885left after every check22 completes too fast or too slow

Rules per survey

Pick the checks for each survey

In the survey's settings, switch on the checks that suit the study. A household study might leave the same-IP check off, because family members share one internet connection.

US-only consumer study

Strict: every check on, and a failed check terminates.

Duplicate IPDeviceCountryVPN

Household study

Same-IP check off, because people at one address often share an IP address. The device check stays on.

Duplicate IPDeviceCountryVPN

Terminate or flag

You choose what a failed check does. Terminate sends the person to the survey's termination message. Flag lets them finish and marks them in the validation summary for you to decide.

Reasons

A reason you can pass on

Because every rejection is written out, you can show a client or a sample supplier exactly why a respondent was removed.

Every failure is recorded in plain words, such as "Another response from the same IP address within 24 hours" or "VPN, proxy or hosting-provider address".

Two layers

Who gets in, and how they answered

FraudCheck decides who gets in. Response quality checks look at how people answered, such as rushing, straight-lining, gibberish and off-topic answers.

FraudCheck

Decides who gets in

As each person opens the survey

  • Duplicate IP
  • Duplicate device
  • Country mismatch
  • VPN and proxy
  • Threat score
  • Interview length limits

Response quality checks

Looks at how they answered

While they answer, and after

  • Speeder
  • Straight-lining
  • Open text
  • Gibberish
  • Profanity
  • Off topic
  • Outlier
  • Video
See the response checks
One place for all quality decisions

VPN or proxy, country and threat-score results appear in the validation summary next to the answer checks, so entry and answer quality are judged together.

Example

Example: a US-only survey shared by open link

A phone brand shares a survey link on social media to reach US adults who bought a new phone this year. Open links attract repeat and out-of-country entries, so the team switches on country mismatch with US as the only allowed country, plus the device check, VPN and proxy, and a minimum length based on their own timed test runs. For the first day they choose Flag, read the reasons in the validation summary and make sure real buyers are not being caught. Then they switch to Terminate for the rest of fieldwork.

On our own panelNothing earned for a security failOn our PocketsFull panel, a member removed by a security check is not paid for that survey, so cheating does not pay.

How members are paid

Security and research standards we work to

  • Cyber Essentials certified
  • ISO 9001, ISO 20252 and ISO 27001
  • ESOMAR Corporate member
  • MRS Company Partner
  • AQR, the Association for Qualitative Research
Our accreditations

FAQ

Questions, answered

Will people in one household be blocked?

Only if you switch on the duplicate IP check, because people at one address often share an IP address. For household studies, leave that check off and rely on the device check.

What does a person who fails a check see?

If the survey is set to Terminate, they see the survey's termination message. If it is set to Flag, they finish as normal and the flag appears in the validation summary.

Does FraudCheck look at the answers themselves?

No. FraudCheck decides who gets in. Response quality checks look at how people answered, such as rushing, straight-lining, gibberish and off-topic answers.

Are panel members paid if a security check removes them?

Not on our PocketsFull panel. A security fail earns the member nothing for that survey.

Set the rules before fieldwork starts

Tell us who the survey is for and how it will be shared. We will help you pick checks that keep fake entries out without turning real people away.

Nearest office: Orlando, FL (corporate office)