Skip to main content
Behavioral Analytics Score Example

​
What is the Behavioral Analytics Score?

The Behavioral Analytics Score (BAS) measures the quality of respondent interaction behaviour to distinguish between natural human input and potentially artificial or automated responses. It helps detect answers generated by AI bots, scripts, or copy-paste actions—enhancing the reliability and authenticity of survey data. Each BAS data point receives a score from 0–100 and a category. When both typing and mouse behaviour are available, ReDem scores them separately and uses the lower score for the data point. If both scores are equal, the typing category is used.

​
High-level checks performed

  • Checks for unusually fast typing that is inconsistent with manual human input.
  • Checks for copy-&-paste patterns (e.g., large blocks inserted with minimal keystroke events).
  • Checks for highly uniform or repetitive keystroke rhythms that differ from typical human variation.
  • Where instrumented, checks for behaviour metadata (e.g., pause durations between keystrokes, editing patterns) that deviate from expected human behaviour.
  • Swipe Typing (Glide Typing): Detects swipe/glide typing on Android and iPhone. Classified as NATURAL_TYPING with a BAS score of 60. The score is intentionally conservative to reduce the impact of potential false negatives on the overall BAS score.
  • Voice Dictation (Speech-to-Text): Detects speech-to-text input on Android and iPhone. Classified as NATURAL_TYPING with a BAS score of 60. The score is intentionally conservative to reduce the impact of potential false negatives on the overall BAS score.

​
Typing categories

Typing is scored when at least 3 KEYSTROKE or COPY_AND_PASTE events are present for the data point.
  • NATURAL_TYPING — typing behaviour looks human-like (score ≥ 50). Also used for detected swipe typing and voice dictation (score 60).
  • UNNATURAL_TYPING — typing behaviour looks automated or scripted (score < 50). Timing, rhythm, and editing signals that look non-human are reflected in this category — there is no separate “unnatural keystrokes” label.
  • COPY_AND_PASTE — the interaction is entirely paste-based (100% of events are COPY_AND_PASTE). Score is 0. Partial paste mixed with keystrokes is scored as typing instead.

​
Mouse Movement and Clicks

When your integration tracks pointer behaviour, BAS evaluates mouse movement and clicks alongside typing. Send MOUSE_MOVEMENT and MOUSE_CLICK events in the same interactionData array as keystrokes. See the Behavior Tracking guide for a sample implementation. How scoring works
  • ReDem scores typing and mouse behaviour separately.
  • The lower score is used for the BAS data point. If both scores are equal, the typing category is used.
  • Typing still requires at least 3 keystroke or copy-paste events. Mouse behaviour can be scored on its own when movement or click events are present.
  • ReDem only uses the last 300 MOUSE_MOVEMENT and last 300 MOUSE_CLICK events per data point.
What mouse behaviour is checked
  • Unusually large or instant pointer jumps
  • Repeated high-speed teleport-style movement
  • Perfectly straight horizontal or vertical movement over longer distances
  • Clicks placed at the exact centre of the target element
  • Bursts of very rapid clicks
Categories
  • NATURAL_MOVEMENT — mouse behaviour looks human-like (score ≥ 50)
  • UNNATURAL_MOVEMENT — mouse behaviour looks automated or scripted (score < 50)
UNNATURAL_TYPING, COPY_AND_PASTE, and UNNATURAL_MOVEMENT can be used as category-based cleaning and exclusion rules. See the Data Cleaning guide.

​
How to use the BAS?

The BAS is designed to be used in real-time, via our API, or through survey software with integrated ReDem functionalities. This quality check is not available via the Quick Import.