Skip to content
Optima docs
ConsoleCampaignsJourneys

Split node

Divide the profiles in a journey between two or more arms by percentage, and optionally let the journey find the winning arm.

A split node sends profiles down different arms by percentage. Use it to compare two subject lines, two offers or two timings, or simply to hold some customers back on a different path. A split can run its shares as you set them, shift traffic toward the arm that converts best, or run a winner test: try the arms on a slice of the audience, then send everyone else down the winner.

Who uses it

Campaign creators. Changing the node needs campaign_editor, or another role your administrator has given access to this area — see Licences and roles.

Fields

Arms

FieldMeaningValid valuesDefault
LabelThe arm's name. It also labels the arm's outgoing edge.Text; unique within the splitNone
WeightThe arm's share of the profiles, as a percentage. All weights add up to 100.Whole number, 1 to 100None

A split needs at least two arms with unique labels, and their weights must add up to 100. The dialog shows Total as a percentage, and Make it 100% scales the weights to fit. Add arm and Remove arm change the number of arms. Each arm needs exactly one outgoing edge carrying its label: draw the edge from the split to the first node of that arm.

Optimization

FieldMeaningValid valuesDefault
OptimizationWhat the split does with the arms over time.None — run the shares as configured, Auto-optimize — keep shifting traffic to the best arm, Winner test — test a slice, then commit to the winnerNone
Winning meansThe goal node that decides which arm is better. The arms must lead to it, and the arm that converts that goal best wins. Required for the two optimizing modes.A Goal node in the same journeyNone

A journey with no goal node cannot optimise: add a goal and point the arms through it first.

Auto-optimize

Shifts traffic toward whichever arm converts the goal best. It never commits, so a losing arm can win back traffic. The shares you set are the starting point, not a fixed allocation.

FieldMeaningValid valuesDefault
Keep exploring (%)The share of traffic spread evenly over all arms for ever, so a losing arm keeps a trickle and can recover. 0 uses the default.0 to 9920
Runs per arm before shiftingUntil every arm has this many runs behind it, the shares above are used unchanged, so one lucky early conversion cannot take the whole audience.Whole number; 0 uses the default100

Winner test

Only the test share of the audience runs the arms now. The rest wait at the split. When the evidence window closes, the arm with the best conversion rate toward the goal wins, and all held profiles flow down it. There is one test per saved version of the journey.

FieldMeaningValid valuesDefault
Test share (%)How much of the audience runs the arms during the test. The rest, shown as the held share, waits for the winner.1 to 9920
Minimum conversions to winAn arm below this cannot be declared the winner. If no arm reaches it, the test ends inconclusive and held profiles use the configured shares.Whole number; 0 uses the default3
Decide afterThe evidence window, measured from the first profile to reach the split. Held profiles wait exactly this long. Required.A number with minutes, hours, days or weeks1 day

A paused campaign does not stop the clock. If the goal itself waits longer than this window for conversions, the dialog warns you: conversions after the window do not count toward the verdict. That only affects the measurement; nobody's message is delayed.

Actions

  • Save puts the arms and settings on the canvas. Save the journey to keep them. Editing and saving the journey starts a new test, because the verdict is per saved version.
  • While a winner test collects, the runs monitor shows a Winner tests card with the status (collecting, decided, inconclusive), the winner, the deadline and the evidence. Decide now concludes it early on the conversions so far, after you confirm. Every held profile then flows at once, and the test cannot be rerun on that version. The verdict counts every run on the saved version, including runs that entered the journey before the first held profile reached the split.
  • The runs monitor also shows a Split distribution: how many runs on the current page went down each arm.

Examples

Plain 50/50. Two arms, a and b, weights 50 and 50, Optimization None. Each arm leads to a send with different copy.

Find the better subject line. Arms short and long, 50 and 50. Winning means a goal node that waits for a purchase for 7 days. Optimization Winner test: Test share 20%, Minimum conversions to win 30, Decide after 7 days. After a week the better arm takes all the remaining 80%.

Keep learning. Auto-optimize with Keep exploring 20% so a weaker arm keeps getting some traffic.

Common problems

  • The editor says a split needs at least two arms or Arm labels must be unique. Add an arm or rename.
  • Split shares must add up to 100%. Use Make it 100%.
  • Pick the goal an optimizing split measures by. Choose a goal node under Winning means; if none exists, add one first.
  • Set the winner test's decide-after window. Decide after is required for a winner test.
  • The test ended inconclusive. No arm reached Minimum conversions to win. Held profiles used the configured shares. Edit and save the journey to test again.
  • Held profiles are waiting. That is the winner test. They wait for the evidence window to close, or for Decide now.

On this page