Survey branching and skip logic: how to design paths that preserve data quality

Design survey branching as a small decision system, not as a collection of page jumps. Give every rule an identifier, define its condition and destination, then test one synthetic respondent for every terminal route.

What are branching and skip logic?

Survey branching routes respondents to different questions or endings based on earlier answers or known attributes. Skip logic is the narrower instruction to bypass questions that do not apply. Display logic changes whether a question appears, while piping inserts a previous answer into later content.

How should a branch be specified?

Record each rule in a table before building it. Use stable question IDs rather than display numbers, because numbers can change during editing.

How do you map all paths?

1. List entry questions and every possible answer.
2. Draw one edge from each answer to its next question or terminal.
3. Mark required fields on each route.
4. Find orphan questions with no incoming edge.
5. Find dead ends with no valid completion state.
6. Create one test profile per terminal route, plus boundary cases for multi-select conditions.

Worked example

A product feedback survey asks whether a respondent has used the product. Users see experience questions; non-users see awareness questions. A third path is needed for respondents who cannot remember. Sending that answer to either user or non-user questions would force inaccurate data. The branch map makes the missing route visible.

What commonly breaks?

• Overlapping rules: two conditions are true and rule priority changes the route.
• Stale references: a question is deleted but a branch still points to it.
• Back-button drift: an earlier answer changes without clearing now-inapplicable data.
• Export mismatch: skipped and unanswered values are indistinguishable.
• Unstable numbering: documentation points to Q7 after editing renumbers it.

How much branching is too much?

Branching is too complex when the team cannot enumerate and test all material routes. Split the survey, reduce conditions or move personalization outside the questionnaire if path count grows faster than the evidence value.