Send people down different paths

A split is a fork. It tests something you already know about this visitor and sends them down one of two paths. The visitor never sees it — they just find themselves on a screen that suits their answers.

Splits are the reason a conversational form can be shorter than a form. Instead of asking everyone everything, you ask each person only what applies to them.

#Add one

#Building the condition

There are two ways to write it, and a button to switch between them.

#Visual mode

Three fields per condition:

Add Condition adds another row when one test is not enough. Underneath, a Generated Expression line shows you what you have built, which is the easiest way to check you have said what you meant.

#Code mode

Switch to Advanced Mode (Code) to type the condition yourself. It is also where you write tests the list does not offer, such as JSON() or $DATASETFILTER.

$AGE > 18 AND $PLAN = GOLD

How a condition is read:

• Put a space on each side of every operator. $AGE > 18 works, $AGE>18 does not.
• Type values without quotes. $CITY = New York matches New York; $CITY = 'New York' looks for the quote marks too.
• Text ignores capitalisation, so US and us match.
• A value starting with $ is another variable. $ECODE = $VALIDATION_CODE compares two values; $PLAN = GOLD compares with the word GOLD.
• Use either AND or OR in one split, not both. Brackets do not group conditions, and a split that mixes the two behaves as if every link were AND. For “A and B, or C”, chain two splits.

Numbers are compared as numbers when the question is an A Number question, and only whole numbers work there. Anything else is compared as text, where 10 comes before 9.

#What you can test

Beyond the answers this flow collected, splits can read these System Variables:

• Flow Points, Flow Parameter and Flow Session ID
• Lead Name, Email, Phone, Stage, Quality, Tags and the other lead fields, when the visitor is a contact you already know
• Assigned First Name and the other details of the team member the conversation is assigned to
• Results that actions stored under an identifier, such as a webhook reply or an AI Prompt answer

In code mode you can also test:

• $DATASETFILTER, how many rows the last Filter DataSet found
• a value inside a JSON reply, with JSON($CUSTOMER[address][city]) = Lisbon

That makes some useful patterns easy:

#Both branches need somewhere to go

Every split has two outputs, labelled True and False on the canvas.

If you genuinely do not need two paths, the False branch can be a single thank-you screen. It still has to exist.

#Chaining splits for more than two outcomes

A split gives you two outcomes. For three or more, put a split on a branch of another split.

Score >= 80 ?
  True  -> "Excellent" screen
  False -> Score >= 50 ?
             True  -> "Good" screen
             False -> "Needs work" screen

Order matters: test the highest threshold first. If you test >= 50 before >= 80, everybody scoring 90 takes the first branch and the second is never reached.

#When a condition seems to be ignored

Work through these:

1. Is the identifier spelled exactly right? $BUDGET and $BUGET are not the same, and a name that matches nothing never passes the test.
2. Does the split come after the question? A split placed before the question it tests has nothing to read.
3. Are you comparing a number as text? If the question is a Text type, 10 may sort as text rather than as a number. Ask numeric questions with A Number.
4. Are you comparing against the label or the value? If your options have a Value set, that is what gets stored, not the label the visitor saw.
5. Did you put quotes round the value, or leave out the spaces? See Code mode.