Data quality

Find responses that deserve a closer look

Response confidence combines several quality signals into one explainable score. Use it to decide where human review is most useful, not to label a respondent or automatically reject their answers.

A review signal, not a verdict

A score of 100 means none of the enabled checks reduced confidence. It does not prove identity, authenticity or attention.

80-100 trusted 40-79 review 0-39 risky

Use confidence to focus your review

Quality checks run quietly while people answer. They can flag responses for review without interrupting the respondent, showing a CAPTCHA or rejecting a submission.

Start with the lowest-confidence responses, inspect the reasons in context and compare them with the respondent source and the answers themselves.

Try asking your AI

“Which responses have low confidence, and what should I review first?”

“Summarize the confidence profile of this survey without treating the score as proof of fraud.”

What the checks look for

Dayalogs uses several kinds of signal so that no single behavior decides the result. The Help deliberately describes these at a high level: they are review aids, not instructions for passing a test.

Response timing

Highlights timing that may be unusual for the survey experience.

Session behavior

Looks for browser-session behavior that may deserve contextual review.

Answer patterns

Looks for mechanical patterns in batteries while treating legitimate non-substantive answers carefully.

Collection patterns

Compares responses when enough fieldwork exists to spot repetition that one response cannot reveal.

Privacy: quality checks can use behavioral and device-derived signals, including protected technical identifiers. Describe their use in the survey privacy notice when appropriate.

Choose which checks count

Open the survey's Confidence tab to enable or disable checks for that fieldwork setup. For example, shared event tablets may produce normal behavior that looks unusual in a survey answered at home.

When settings change, Dayalogs recalculates the score without losing the recorded quality signals. Disabled checks remain visible in a muted state so you can still understand the underlying fieldwork.

Try asking your AI

“This survey ran on shared event tablets. Disable the timing check and recalculate confidence.”

Read individual and survey-level results

The Responses tab shows a compact confidence bar for each completed response. The Confidence tab shows the overall score, responses sampled and the result for each enabled check.

Response #24
92%
Response #25
58%
Response #26
18%

Use the same active checks when filtering or exporting so the review threshold means the same thing throughout the project.

Try asking your AI

“Export completed responses with confidence of 80 or higher.”

What confidence cannot tell you

Confidence is probabilistic. It cannot prove who answered, guarantee that a response is genuine or replace good sampling and fieldwork controls. Distributed or low-volume problems may not create a visible pattern, and comparison checks need enough responses before they become useful.

Keep the score alongside your research judgement. Review the source, context and answers before deciding whether a response should remain in the analysis.