Skip to content

How to Analyze Respondent Behavior and Data Quality ​

INFO

Behavioral Analytics uses client-side paradata (interaction timing, answer revisions, paste bursts, and tab focus) to assess data quality and survey UX friction. Researchers can identify confusing questions, detect low-effort submissions, and audit or exclude problematic responses without compromising privacy.

Behavioral analytics dashboard overview showing KPIs, quality charts, friction plot, and respondent log
The Behavioral Analytics dashboard provides holistic visibility into respondent effort, friction, and submission integrity.

Before you begin: enable behavioral tracking ​

The Behavior tab only has data for surveys that enabled tracking before respondents took them. Turn it on once per survey:

  1. Open the survey and switch on Advanced Mode.
  2. Open Share → Behavioral Tracking.
  3. Turn on Activate Behavioral Tracking.
  4. To analyze abandonment as well as completed sessions, turn on Monitor Drop Off. This also turns tracking on; turning tracking off turns drop-off monitoring off.
  5. Publish the survey.

Tracking applies to sessions that start after it is enabled. If the Behavior tab is empty, check this setting first — and note that behavioral tracking is available on some plans only. See Behavioral Tracking Reference for what is recorded and the plan limitation.

1. Opening the Behavioral Analytics View ​

To access behavioral analytics:

  1. Open your survey in the management console and click the Analyze tab in the main navigation.
  2. In the secondary sub-navigation, click Behavior.
  3. The dashboard loads behavioral paradata gathered from respondents during their survey sessions.
Navigation to the Behavior tab in the Analyze navigation bar
Click the Behavior tab within the survey Analyze navigation bar.

Privacy and Ethical Tracking

Behavioral analytics is non-invasive: the platform does not log keystrokes, record screen video, or capture biometric data. It measures event timings (active vs. idle duration, tab focus changes, choice revisions, and paste events) to evaluate aggregate survey health and question quality while respecting respondent privacy.

2. Reading the Behavioral KPI Summary Strip ​

The top KPI strip provides a high-level summary of survey engagement and data integrity:

Behavioral KPI strip displaying active time, confidence rate, drop-off rate, and burst rate
The KPI strip summarizes session duration, cognitive attention, data quality, and attrition at a glance.
  • Median active time: The typical time respondents spend actively interacting with questions (excluding idle time when the browser is inactive).
  • Mean focus ratio: The average share of session time respondents were actively focused on the survey tab.
  • High confidence (≥ 70): The percentage of scored responses whose Confidence Index is 70 or higher.
  • Drop-off rate: The percentage of started sessions that were discontinued before completion.
  • Paste / burst rate: The proportion of tracked responses where long text answers were pasted or entered in single high-velocity bursts.

Hover, focus, or tap the info icon in the corner of each tile to read what the metric measures.

Selection summary displaying total and filtered respondent count
The dataset summary indicates total submissions and how many records match the active filters.

3. Assessing Data Quality: Confidence Index and Quality Signals ​

The dashboard evaluates submission authenticity through two interactive visual charts:

Confidence Index Distribution ​

The Confidence Index card displays a score histogram (from 0 to 100) classifying responses based on behavioral engagement:

Confidence Index score distribution histogram
The Confidence Index distribution classifies responses from high-confidence human engagement to potential speeders or bots.
  • High Confidence (≥ 70): Indicates steady reading dwell times, high focus ratios, and deliberate interaction.
  • Low Confidence (< 50): Highlights responses exhibiting rapid click-through speeding, high idle gaps, or paste bursts.

Rubric and Theory

For the exact mathematical penalties and heuristic scoring formulas, see Confidence Index Scoring Rubric. For the scientific foundation and satisficing research, see Understanding Respondent Behavior and Data Quality.

Quality Signals Breakdown ​

The Quality Signals chart categorizes responses into behavioral indicator flags:

  • Attentive: Complete sessions meeting human reading dwell thresholds.
  • Revised choices: Respondents who altered their initial choice selection, indicating deliberation.
  • Paste / burst: Instantaneous text entry flagging copy-pasting or automated script injection.
  • Too fast: Submissions completed faster than human reading limits allow (speeders).
  • Zero edits: Direct single-pass submissions without revisions.
  • Dropped off: Abandoned partial sessions.
Quality signals bar chart showing counts for attentive, revised choices, paste burst, and speeders
Quality signal flags categorize specific behavioral indicators across all completed and partial submissions.

Clicking any score bucket or signal bar immediately filters the entire dashboard to that respondent cohort.

4. Diagnosing Drop-Off Points ​

When respondents abandon a survey, identifying the exact exit location is critical for UX improvement. Use the segmented toggle above the drop-off card to switch between analysis levels:

Drop-Off by Page ​

Identifies the overall survey section or page where abandonment occurred:

Drop-off distribution segmented by survey page
Page-level drop-off distribution highlights which survey stage experiences respondent attrition.

Drop-Off by Question ​

Pinpoints the exact question where respondents halted their participation:

Drop-off distribution segmented by specific question
Question-level drop-off isolates the exact question where respondents discontinued their session.

Remediation Strategy

If a specific question exhibits a drop-off spike, consider whether it asks for sensitive information, contains confusing instructions, or is excessively long.

5. Per-Question Attention and Cognitive Friction ​

The Per-Question Attention & Friction card analyzes response latency across every question to detect cognitive stumbling blocks:

Median Duration Mode ​

Displays the median dwell time per question against the overall survey median:

  • Questions with dwell times at least twice the survey median are automatically flagged as Hesitation Hotspots.
  • Hotspots indicate questions requiring disproportionate cognitive effort, complex phrasing, or confusing layout.
Per-question attention and friction chart in median duration mode
The median friction chart flags hesitation hotspots requiring disproportionate cognitive effort.

Boxplot Mode ​

Use the Median + IQR / Box plot toggle in the card header and select Box plot to visualize dispersion:

  • Displays the median, the interquartile range (IQR, from the first quartile to the third quartile), and minimum/maximum whiskers.
  • Reveals answer latency spread and distribution skewness, helping distinguish questions that trouble all respondents from questions that only confuse a subset.
  • Dots beyond the whiskers are outliers.
Per-question friction chart in boxplot mode showing IQR and whiskers
Box plot mode displays response time dispersion, interquartile ranges (IQR), and outliers per question.

Chart Options and Export ​

The card header also holds the sort toggle. You can:

  • Switch between Median + IQR and Box plot representations.
  • Switch sort order between Slowest first (prioritizing friction) and Survey order (following question order).
  • Hover or focus a toggle to read a tooltip explaining what its representation shows.
  • Export the friction visualization as high-resolution PNG or SVG from the card menu (three dots).
Friction chart options menu showing PNG and SVG export options
The card menu exports the friction chart; the representation and sort toggles sit in the card header.

6. Interactive Behavioral Cross-Filtering ​

Every behavioral chart mark serves as a dynamic cross-filter. Clicking any element filters the entire behavioral view to that cohort:

  1. Click any quality signal bar (e.g. Paste / burst or Too fast) or a Confidence Index score bucket.

  2. The active filter chip appears in the summary bar.

    Active filter chip reflecting behavioral criteria
    Active behavioral filter chips isolate subsets (e.g. low-confidence or high-friction respondents).
  3. The entire KPI strip and respondent log immediately recalculate to reflect only that cohort:

    Dynamically updated KPI strip reflecting filtered respondent subset
    All KPI metrics update instantly to reflect the characteristics of the filtered cohort.
  4. To reset filters, click the × on the filter chip or click Reset all filters in the toolbar.

7. Auditing Responses in the Behavioral Respondent Log ​

The tabular Behavioral Respondent Log below the charts provides an auditable row-by-row breakdown of each session:

Behavioral respondent log grid with per-respondent behavioral metrics
The Respondent Log provides an auditable breakdown of timing, edits, and quality signals per response.

Columns include:

  • Checkbox: Selection control for bulk exclusion.
  • Submitted: Submission date and time.
  • Status: Completed (Submitted) or partial (Drop-off).
  • Confidence: Color-coded Confidence score badge (green for scores of 70 or more, amber for 40 to 69, red below 40).
  • Quality Signals: Active flags assigned to the respondent.
  • Active Time & Idle Time: Clean separation of active interaction versus idle wait time.
  • Focus %: Proportion of session time spent inside the survey tab.
  • Choice edits & Paste bursts: Raw event counts.
  • Slowest question: The specific question where this respondent spent the most time.

You can sort by any column header or use the column filter inputs to isolate specific patterns (for example a confidence score below 50, or an active time below 10 seconds).

8. Inspecting Individual Respondent Timelines ​

To analyze a specific respondent's trajectory through the survey:

  1. Click any row in the Respondent Log to open the Respondent Inspector dialog.
  2. The inspector displays:
    • Overall session metrics (active time, idle time, focus %, speed penalties, and interaction scores).
    • Question-by-question horizontal bar timeline displaying dwell time per question.
    • Distinct highlight for the respondent's slowest question and drop-off point.
Respondent inspector dialog showing question-by-question timing breakdown
The respondent inspector details question-level attention and identifies individual slowest questions.

9. Excluding and Restoring Low-Quality Responses ​

When suspicious, automated, or invalid responses are identified, researchers can exclude them from the analytical dataset while preserving complete audit compliance:

Excluding Responses ​

  1. In the Behavioral Respondent Log, select the checkboxes next to the responses you want to remove (this requires edit rights on the survey). The header checkbox selects every row in the current filter.

  2. The table header shows N selected; click Remove from dataset…:

    Respondent log selection bar showing one selected respondent and the Remove from dataset button
    Selected rows are counted above the log, next to the Remove from dataset… action.
  3. In the exclusion dialog, select an auditable Reason:

    • Test data: pre-launch checks, internal trials, and QA submissions.
    • Likely bot: automated submissions flagged by instantaneous text entry or impossible completion speeds.
    • Duplicate: multiple entries from the same respondent or device.
    • Other: any other reason; a Note is required and explains what it was.
  4. Add a Note when needed and click Remove.

    Remove responses dialog with the reason selector and note field
    The removal dialog records a reason and, for Other, a required explanatory note.

Excluded responses are immediately omitted from all Analyze charts, behavioral KPIs, and filtered browser exports. The same removal tool is available on the Analyze tab: turn on Advanced and use the response grid.

Archival and Audit Preservation

Exclusion never permanently deletes survey data. The original submission remains safely stored in the response database, the raw Progress tab, and the compliance archive.

Reviewing and Restoring Excluded Responses ​

  1. When excluded responses exist, an Excluded (N) button appears in the toolbar showing the count.
  2. Click the button to open the list of excluded records. Each entry shows the exclusion reason, note, submission date, and the operator and timestamp of the exclusion.
  3. Click Restore next to any record to reinstate it into the analytical dataset.

10. Advanced Behavioral Statistics ​

Toggle on Advanced in the top toolbar to display a live statistical summary strip below each behavioral card:

  • Numeric Metrics (Confidence Index): Displays n, Missing, Mean, SD, Median, IQR, and Range.
  • Categorical Metrics (Quality Signals): Displays n, Missing, Top signal with percentage share, and total Categories.
Advanced mode statistical summary strip below the Confidence Index chart
Advanced mode adds mean, standard deviation, median, IQR, and range to the Confidence Index card.
Advanced mode statistical summary strip below the quality signals chart
Quality Signals gains a top-signal share and category count below its chart.

See Behavioral Analytics Reference for complete metric specifications and formulas.

11. Exporting Behavioral Datasets ​

To analyze timing paradata or train statistical models outside the platform:

  1. Click the Export button in the behavioral toolbar.
  2. Tick I understand my responsibilities to acknowledge the data privacy notice.
  3. Select Download from this browser to download the behavioral dataset matching your active filters.
Top of the export dialog on the Behavior tab with privacy warning and download method
The Behavior tab uses the shared export dialog; browser downloads honor the active behavioral filters.

The browser download generates a row-by-row dataset with all Behavioral Respondent Log indicators and question-level active durations.

Detailed Export Options

For complete instructions on configuring metadata columns, raw codes vs. humanized labels, multi-value separators, or downloading the full server compliance archive, see How to Export Survey Data.