​
What is the Grid-Question Score?

The GQS helps assess data quality in grid questions (also known as matrix questions). It uses machine learning and a defined set of pattern-detection rules to evaluate how respondents interact with grid items.

​
How is the Grid-Question Score calculated?

Each grid question is scored individually based on the respondent’s answer sequence. We compare that sequence to known low-quality response patterns; the closer the match, the lower the grid’s score. Finally, all grid-level scores are combined into one overall GQS per respondent. Low-quality patterns we detect:
  • Straightlining
    Selecting the same option across all items. Labeled “Straightliner.”
  • Partial Straightlining
    Mostly the same option with only a few switches between adjacent items. If fewer than 35% of adjacent transitions are changes, we label it “Partial Straightliner.”
  • Pattern clicking
    Mechanical rhythms that ignore item content, e.g. alternating (1-2-1-2…), stair-stepping (1-2-3-4…), zig-zag/mirrored patterns (1-2-3-2-1…), or repeating cycles. The more regular and repeated the pattern, the lower the score.
To detect response patterns beyond straightlining, grid question data must be submitted to the API in the original display order shown to the respondent.
Grid-Question Score Example

​
What Patterns Does the GQS Detect?

ReDem currently analyzes the following behavioral patterns in grid responses: Grid-Question Score patterns