How we check quality
Every response on Tayqun is checked before it counts. Researchers get data they can defend, honest participants get paid, and people trying to game the system find it isn't worth the effort.
What we check on every response
Checks run at four points: who is answering, how they answer, what they answer, and again after the response is in.
- 01
Who is answering
One real person, one account, in the right place.
Identity and one account per person
Every participant signs up with a phone number, and one person can hold only one account. Device and network signals show us when several accounts come from the same place, and those accounts are reviewed together.
Location consistency
We compare the country a participant says they live in with where they connect from and the time zone on their device. When these disagree with each other, or with the countries a study targets, the response gets a closer look.
- 02
How they answer
Real reading, real attention.
Answering speed, section by section
We time each section against how long it takes to read, not only the survey as a whole. Rushing through one block of questions is caught even when the total time looks normal.
Attention checks and hidden trap questions
Studies can include questions with one clearly correct answer, and questions that people never see but automated tools do. Missing the first or answering the second is a strong sign that no attentive person was behind the response.
- 03
What they answered
Answers that are original and consistent.
Duplicate and copied answers
Responses are compared with one another. Identical or near-identical answer sets, repeated written answers and the same rating on every item are all flagged.
Retried screening questions
Screening questions decide who is eligible. If someone is screened out and comes back with different answers to get in, we see it, and the later attempt is held.
Household consistency in parent-report studies
When one parent answers for several children, we check that the household holds together: a plausible number of children, sensible timing between the siblings' responses, and sibling answers that aren't copies of each other.
- 04
After the response is in
The study is checked as a whole, and people make the call.
Automatic re-check of the whole study
Some patterns only appear across many responses. Each time new responses arrive, the whole study is checked again, so a response accepted early can still be held later if it turns out to be part of a pattern.
Human review
Automatic checks raise a flag; they don't have the final say. A member of our team reviews held responses before any are excluded, and researchers see every flag on their own quality dashboard.
Why we don't publish the exact rules
Limits and scoring are tuned over time, and publishing them would only help someone work around them. Honest participants don't need to know them: answering carefully is enough.
Where a response ends up
After the checks, every response lands in one of three places.
Clean
Passed every check. It counts as normal.
- Researcher charged
- Yes
- Participant paid
- Yes
Accepted with a note
A minor sign was noticed, but not enough to hold the response. You see the note and decide whether it matters for your analysis.
- Researcher charged
- Yes
- Participant paid
- Yes
Held for review
Set aside for a person to look at. If it's cleared, it counts as normal. If not, it's excluded.
- Researcher charged
- No, while held
- Participant paid
- No, while held
The money rules
Held means unpaid and uncharged
A held response is never charged to the researcher and never paid to the participant unless a reviewer clears it.
Excluded responses are refunded
If a response you were charged for is excluded later, for example after a study-wide re-check, the charge goes back to your account.
Rewards are reversed
If a response that was already paid turns out not to be genuine, the participant's reward for it is taken back.
Repeat offenders are blocked
Accounts that keep producing excluded responses lose access to studies, and linked accounts go with them.
What researchers receive
Quality work is only useful if you can see it. Every study comes with the evidence, in a format your analysis can use.
An Excel export with a column guide
One row per response and one clearly named column per question, plus a codebook sheet that explains each column: the question wording, the answer codes and the scale it belongs to.
Quality notes on every response
Each row carries its quality status and the reason behind any note or hold, so you can run your analysis with or without flagged cases.
Anonymised household IDs
In parent-report studies, children from the same family share an anonymous household ID and a child number, so you can account for siblings without knowing who anyone is.
| ResponseID | HouseholdID | ChildNumber | Quality | QualityNote |
|---|---|---|---|---|
| R-1048 | HH-0137 | 1 | Clean | - |
| R-1049 | HH-0137 | 2 | Noted | Fast section 3 |
| R-1050 | HH-0152 | 1 | Clean | - |
For participants
Most people never notice these checks. They answer naturally and get paid. Three things keep it that way.
- 1
Answer honestly, at your own pace
Read each question and give your real answer. There's no reward for speed, and rushing is the most common reason a response is held.
- 2
One person, one account
Don't open extra accounts or share yours. Linked accounts are reviewed together and can be closed together.
- 3
Rewards are paid after the checks
Most responses clear quickly. If yours is held, a person reviews it, and you can ask us to look again if you think we got it wrong.
Questions we're often asked
Can an automatic check reject a response on its own?
No. Automatic checks can hold a response, but excluding it is a decision made by a person on our team after looking at the whole response.
Do researchers pay for responses that fail the checks?
No. Held responses aren't charged, and if a response you were already charged for is excluded later, the charge is refunded to your account.
Why don't you publish the exact thresholds?
Because publishing them would help people tune fake responses to pass. The checks are described here so you know what we look at; the exact limits stay private and are adjusted over time.
I answered honestly but my response was held. What now?
Being held is not an accusation. A person reviews it, and if it reflects genuine effort, the full reward is released. If you disagree with the outcome, contact support and we'll look again.
Can I see why a particular response was flagged?
Yes. Each response in your dashboard and in your Excel export carries its quality status and a plain-language reason for any note or hold.
What data do these checks use?
Answers and their timing, network and device signals, device time zone and profile details. They're used only to check quality and prevent fraud, as described in our privacy policy.
Running an academic study?
Verified participants, consent in your approved wording and an analysis-ready data file. Get a proposal in one business day.