Skip to main content
The coherence score is an AI-driven metric that evaluates the plausibility and logical consistency of responses throughout an entire interview, from screening questions to socio-demographics. This highly robust quality check detects contradictions and inconsistencies, making manipulation nearly impossible. Unlike traditional trap questions—which fraudsters and AI bots can easily identify—the Coherence Score is a game-changer in fraud detection. It works seamlessly within regular questionnaires, identifying inconsistencies without relying on special questions. For example, it can flag contradictions, like a respondent claiming regular subway use while living in a city without one. Additionally, the Coherence Score minimizes false positives by avoiding the disqualification of respondents who may have misclicked or momentarily lost focus but otherwise performed acceptably. This approach not only enhances fraud detection but also ensures more accurate and fair assessments.

​
High-level checks performed

This check looks at, for example:
  • Whether answers that should logically align (e.g., travel mode, days stayed, budget) are consistent with one another.
  • Whether the pattern and combination of responses across different questions fit a plausible scenario.
  • Whether any answers appear contradictory (for example: “I never use the app” in one section and “I use it weekly” in another).
  • These checks are intended to flag inconsistent or implausible response sets.

​
What the Coherence Score Includes:

  • Explanation of the Rating: Provides a summary of the overall quality of the interview.
  • Describes how consistent or contradictory the respondent’s answers are across questions.
  • Highlights key factors that influenced the rating (e.g. logical flow, alignment of answers).
  • List of Incoherent Questions: Identifies specific questions where the respondent’s answers were inconsistent or contradictory.
  • Points out answers that conflict with earlier responses or are implausible in the context of the interview.
  • Offers valuable insights into areas of potential concern, helping to flag data reliability issues.
If Coherence Score cannot be calculated (for example a timeout or the check is unavailable), the score is shown as a dash instead of a number.

​
Survey Description

The survey description is extra context you give the Coherence Score (CHS) before scoring. It is not part of the respondent’s answers. ReDem passes it to the AI as analyst notes so the model can interpret the interview with the right background and apply any extra rules you set. You can add it in Quick Import as Survey Description & Exclusion Rules, or via the API as surveyDescription on the CHS data point. If your organization requires it, the field is mandatory on import. When provided, it must be between 20 and 500 characters.

​
How it works

  1. Write a short description of the survey and any extra scoring rules.
  2. ReDem includes that text when each interview is scored.
  3. The AI uses it as background only. It does not treat anything in the description as something the respondent said or selected.
AI knowledge cutoff: The AI model has a knowledge cutoff of May 18, 2026. It may not recognize products, brands, venues, or events released after this date. If the survey covers anything newer, include it in the survey description so the Coherence Score does not treat unfamiliar names as implausible.

​
What you can add

Use the survey description for:
  • Survey topic and purpose: what the study is about and what it aims to measure
  • When it was conducted: the fieldwork date or period
  • Recent context the model may not know: newly opened facilities, product launches, political events, or market changes after May 18, 2026
  • Custom exclusion rules: extra scoring rules, for example excluding respondents who choose a non-existing brand or a specific answer
Provide as much relevant detail as the character limit allows. Clear, specific context leads to more accurate coherence assessments.

​
Example: Survey about a recently opened mall

This survey evaluates shopping experiences at the Grand Central Mall, a new shopping center that opened in downtown Springfield in June 2026. The mall features over 150 stores including major retailers, restaurants, a cinema complex, and an entertainment area. The survey aims to understand visitor experiences, shopping patterns, and satisfaction with the new facility.

​
Example: Exclusion rule

Exclude respondents who select a brand that does not exist in this market. The survey was conducted in June 2026 and covers awareness of brands launched after May 2026.

​
Use of GPT for Coherence-Score

ReDem uses GPT hosted on Microsoft Azure, deployed in the EU region to ensure full GDPR compliance when calculating the Coherence Score.

​
Individual Responses & Anonymity

Each interview is sent to Azure with a fully anonymized ID. Only the individual interview is transmitted per API request—entire survey datasets are never shared.

​
Exclusive Interaction with Azure

Only ReDem communicates with Azure. No information about the origin or source of the data is disclosed to Microsoft.

​
Data Residency

All data related to the Coherence Score is transmitted, stored, and processed exclusively within the EU.