Articles on: Funnel Building

AI Nodes

AI nodes are rotators that learn.


They split traffic between their outgoing connections like a normal rotator node, then use performance data to shift traffic toward the paths that do better.


Here a rotator sends all its traffic into an AI node, which splits it between three lander groups. With no data yet, each group gets 33.33%:


AI node in a funnel routing to three lander groups


Use them when you want FunnelFlux to optimize routing for you, instead of adjusting rotator weights by hand.


To open AI Node Settings, hover over the node and click the edit button, or double-click it. The settings are split across three tabs: General Settings, AI Node Settings and Advanced Settings.


General Settings

Current Optimised Values [1] shows the weights the node is using right now. Until it has enough data, it tells you no optimised values are available yet.


Default Rotation Weights [2] are what the node uses when it has no up-to-date routing data, for example at launch or when traffic is low. They work like rotator weights: relative numbers, with the percentage shown beside each one. Set them to the split you would want at the start of a test.


AI node General Settings with current optimised values and default rotation weights


AI Node Settings

This tab controls what the node optimizes for and how it reads your data.


AI Node Settings tab with optimisation goal, lookback window, update frequency, segmentation and routing goal


Optimisation Goal [1] has two parts:

  • Mode: Revenue per View or Conversion Rate.
  • Target Event: Conversions or one of your custom events.


Revenue per view uses total revenue across all events, but event counts still feed into statistical confidence. Pick the target event that has the biggest impact on revenue.


Lookback Window [2] is how much past data each calculation uses. A 24 hour window means each update looks at the last day of data.


Update Frequency [3] is how often FunnelFlux refreshes the reports behind the weights. Some frequencies are not available with some lookback windows.


Segment By [4] controls how routing data is split. By default it is split by the visitor's traffic source. Higher plans can segment further. If a segment does not have enough data, the node uses the funnel or traffic source average instead.


Routing Goal [5] decides what the node is choosing between:

  • Best Next Node: the default. It routes to the best-performing connected node.
  • Best Next Page: it picks the best page inside the connected nodes.
  • Best Next Node > Best Page: it picks the best node, then the best page inside it.


The last two only work when the AI node connects to page group nodes and nothing else.


Which lookback windows, update frequencies and segments you can pick depends on your plan.


Advanced Settings

This tab limits how hard the node pushes traffic around.


AI node Advanced Settings with stickiness, sensitivity, cap and minimum percentage


Stickiness Mode [1] has the same options as on a rotator: None, Sticky and Anti-Affinity. With Sticky, a visitor stays on the destination they were first sent to, even as the AI changes the weights for new visitors.


Sensitivity factor [2] controls how aggressively the node favours the best performers. Think of it as statistical confidence. Only move it towards More Sensitive when you have plenty of conversions, for example 50+ in the lookback window.


Cap Variant Percentage [3] sets the most traffic any single destination can get. Turn it on if no path should ever take 100% of traffic, however well it performs.


Minimum percentage per variant [4] makes every destination keep at least this share of traffic. Use it with the cap so that no variant takes over while you are still testing.


What happens before there is enough data

An AI node needs data before it can optimize anything.


Until it has enough, it behaves like a normal rotator and uses your Default Rotation Weights. As data comes in, it builds routing profiles and Current Optimised Values starts showing the weights it uses.


This matters: AI nodes are not magic on day one. They need signal.


When to use AI nodes

Use AI nodes when:

  • you have multiple destinations to optimize between
  • conversion or revenue outcomes differ between paths
  • you have enough traffic for learning to matter
  • you want automated optimization instead of manual weight changes


Do not use AI nodes for tiny tests with very little traffic. A simple rotator is easier to reason about until there is enough data.


Add-on status

AI nodes are a paid add-on feature.


For more on how the optimization works, see Using AI Routing Nodes.

Updated on: 13/10/2026

Was this article helpful?

Share your feedback

Cancel

Thank you!