Conditional logic

When to use conditional logic

Use it whenever a question only makes sense to part of your audience. The classic cases are follow-ups ("you said you were dissatisfied — what went wrong?"), role-specific sections (tenants get one block, landlords another), and detail questions that would be noise for everyone who answered "no" to the question above.

The gain is not tidiness, it is length. A respondent judges a survey by the number of questions they are asked, not the number it contains. An 87-question consultation with good logic can present as a dozen questions to someone most of it does not concern, and that difference shows up in completion rate.

How a rule is built

A rule has three parts: the earlier question it reads, a comparison, and a value. Several conditions can sit in one rule, and the rule declares whether it fires on any of them (an OR) or on all of them (an AND).

In NumoForms the rule lives on the question you want to control, not on the question being tested. That is worth knowing because it changes how you think about design: ten follow-ups that all depend on the same trigger are ten independent one-line rules, not a tree. Deleting one of them affects nothing else, and nobody has to trace a diagram to work out why a question appeared.

A worked example

A council is consulting on a proposed low-traffic neighbourhood.

1. Q1: "Do you live inside the proposed zone?" — Yes / No / I work there but do not live there.
2. Q2: "Which street do you live on?" — shown when Q1 any of: Yes.
3. Q3: "How do you usually travel to work?" — shown when Q1 any of: Yes, I work there but do not live there.
4. Q4: "Would a resident permit change your view?" — shown when Q1 all of: Yes, and Q3 is "By car".

Somebody who answers "No" sees one question and a comment box. Somebody who lives in the zone and drives sees four. Both are answering the same survey.

Common mistakes

• Hiding the exit. Logic that routes people past everything should still leave them somewhere to say something. A dead end reads as a broken form.
• Confusing "not asked" with "declined". A hidden question is blank in the export, and that blank is not a refusal. Analyse the two separately or you will understate agreement.
• Testing a question the respondent has not reached. A rule can only read answers given earlier in the flow. Pointing one backwards is the most common reason a question "never shows".
• Assuming answers survive. If a respondent goes back and changes the trigger answer, the follow-ups fold away and what they typed into them is discarded. That is correct behaviour, but it surprises people testing their own survey.

In NumoForms

Logic works identically across all three layouts — classic, conversational and paged — and interacts with two other things worth knowing. Scoring counts only the questions a respondent actually saw, and bands are worked out as a percentage of the points that were available to them, so branching past a section never pushes a score down. And piping degrades safely: a token pointing at a hidden question renders as "your answer", not as raw text.

The full mechanics, including how any/all matching behaves and what appears in the CSV, are on the conditional logic feature page.