Skip to main content

Dispositions (Informative)

A disposition is the final category DQC assigns to each response — for example Qualified Complete, OSQ, or Duplicate. You do not send dispositions. DQC derives them from the status you send plus the quality signals (device score, data trust score, persona, duplicate flag, and any failures) already collected. This page is a reference so you know how your data maps.

Full taxonomy

DQC uses nine dispositions across three categories. For the complete definitions, see Dispositions — full taxonomy, the platform-wide reference outside the Decipher section.


📌 The decision map

DQC maps what the response reports — its status, terminationReason, markers and quality signals. Read the table top to bottom: the first row that matches decides the disposition.

RuleDecipher statusSignal→ DQC disposition
1anya terminationReason of duplicateDuplicate
1Terminateda dqc_duplicate_termination markerDuplicate
2anythe Toolbox never evaluated the responsesee below
3OverquotaOverquota
4Partial (answered ≥1 question, then left)Abandon
5Qualifiedany quality check trippedFlagged Complete
5Qualifiedno quality check trippedQualified Complete
6Terminateddevice score of 10 or belowOSQ
6Terminateda DQC device score, data trust score or persona terminationOSQ
6Terminatedno terminationReason set, but a dqc_*_termination marker presentOSQ
6Terminatedanything else — your own screenerDid Not Qualify

The last row is the one most responses land on. Any <term> of your own — an age screener, a quota condition, a knowledge question — reports as Terminated with an empty terminationReason and no DQC marker, and a respondent screened out by your criteria did not qualify.

OSQ covers the three quality terminations DQC triggers — device score, data trust score and persona — plus any terminate carrying a device score of 10 or below, whoever ended the survey. The fourth DQC termination, duplicate, has a disposition of its own.

Two rows that are easy to misread

A duplicate flag is not a Duplicate. isDuplicate only means "we have seen this participant before". It becomes the Duplicate disposition only when the response reports a duplicate termination — a terminationReason of duplicate, or a dqc_duplicate_termination marker on a Terminated response. A repeat participant who completes normally, reporting no duplicate termination, is a Flagged Complete; one screened out by your age question is a Did Not Qualify.

A respondent who finished is never OSQ. OSQ is decided only on a Terminated response — someone who completes the interview with a low device score gets a DQC, Fraud failure and a Flagged Complete instead. The reverse does not hold: a Terminated response with a device score of 10 or below is OSQ even when your own screener ended the survey.

In both cases DQC maps what the response reports, not what happened in the interview.

When several signals apply

A response can trip more than one signal — a duplicate that also has a low device score, for example. The rules are applied in the order of the table above and the first match wins. Whichever rule decides the disposition, every failure is still stored, so nothing is lost.

What counts as "a quality check tripped"

For a Qualified response, any of these makes it a Flagged Complete instead of a Qualified Complete:

  • a custom failure you sent — honeyPot, trapQuestion, openEndQuestion, other
  • a device score of 10 or below
  • the duplicate flag, even when no termination fired
  • a keyboard masher persona

A response with none of them is a Qualified Complete.


🔁 When the Toolbox couldn't evaluate a response

Sometimes the Toolbox never gets to score a respondent — the submission arrives too quickly, the request is blocked, or the wrapper is tampered with. Those responses reach DQC with a placeholder message in place of the participant id, scores and country.

DQC does not read the placeholder score as a failing score. A score of 0 there means "never measured", not "failed". With the device branch suppressed, the status decides — unless a duplicate termination was reported, which still outranks it:

Decipher status→ DQC disposition
PartialAbandon
QualifiedFlagged Complete, with an Other failure
TerminatedDid Not Qualify
OverquotaOverquota

The Other failure is attached so the flag carries a reason — the response is flagged because it could not be evaluated, not because a specific check failed.

These responses still need to identify their survey

A response that carries neither decipherSurveyIdentifier nor a usable surveyId cannot be attributed to a survey, and DQC holds it back rather than filing it under a placeholder name. Both values are sent automatically by the current integration — see Which survey the response came from — so this only affects surveys running an older version of the script.


How the signals get there

  • Status (Overquota / Qualified / Terminated / Partial) comes from your Decipher survey markers — see How status is derived.
  • dqcFraud / dqcDuplicate are computed by DQC from the device/trust/persona/duplicate data the Toolbox sends — see Automatic and custom failures. You can optionally act on them with Termination Scripts.
  • terminationReason identifies which DQC termination fired, and is empty for your own terminations. On a Terminated response it is one of three things that produce an OSQ — the others are a device score of 10 or below, and a dqc_*_termination marker with no reason set.
  • markers carries every marker on the response, your own screener conditions included. DQC reads it as the fallback when terminationReason is empty.
  • Custom failures you send (honeyPot, trapQuestion, …) flag the response as a Flagged Complete.

✅ Summary

  • Dispositions are computed by DQC — you don't send them.
  • Your status + quality signals (scores, duplicate, failures) determine the disposition via the map above, first match wins.
  • A terminated response with no DQC signal at all is a Did Not Qualify — that covers your own screeners.
  • OSQ is only ever assigned to a terminated response; a completed one is flagged instead.
  • All failures are stored even when a response maps to a single disposition.
  • See Dispositions — full taxonomy for the definitions of all nine.