Node Reference

Every node on a canvas has a type, an optional display name, and a set of configuration fields specific to that type. This page describes every node type, its configuration, and when to use it.

Entry nodes

Entry nodes produce the initial set of entities that flows through the canvas. A canvas must have at least one entry node. Entry nodes have no incoming connections — they are always the first node in a path.

Criteria

A Criteria node defines an inline filter tree, identical to the filter tree used by the standard audience builder. It works as both an entry node (when it has no upstream connection) and as a mid-pipeline filter (when it has one or more upstream connections — the criteria is AND-applied to the incoming set).

When to use as an entry node: When the root conditions are specific to this canvas and you want to define them in one place. You can have multiple entry Criteria nodes on a single canvas — for example, to produce two sets that you later union with an Or node.

When to use as a mid-pipeline filter: When you want to progressively narrow a set without starting a new branch — for example, after an upstream Criteria that selects all customers, add a downstream Criteria that further restricts to customers in a specific region. The compiler ANDs the upstream population with this node’s filter.

Example (entry): A Criteria node that admits users who signed up in the last 90 days and have at least one completed order.

Example (mid-pipeline): 250,000 customers from an upstream Criteria → downstream Criteria (region = “EU”) → 68,000 EU customers.

Audience Reference

An Audience Reference node pulls an existing standalone audience into the canvas as an entry set. The referenced audience’s full membership at query time becomes the input to downstream nodes.

When to use: Use an Audience Reference node when you want to build on top of an audience you already maintain — for example, your “All active subscribers” audience that the data team keeps up to date.

Note: Canvas-owned audiences appear with a “From canvas: <canvas name>” chip in other audience pickers (such as the filter-condition picker on the standard builder). However, they do not appear in the Audience Reference node picker within a canvas. To combine two canvas outputs, connect their upstream branches before the Output nodes rather than referencing one from another canvas.

Set-operation nodes

These nodes combine sets flowing through the graph. To narrow a single set with additional rules, use a downstream Criteria node (see above) — it doubles as a mid-pipeline filter.

And

An And node emits the intersection of all its upstream inputs — entities that appear in every connected upstream set.

When to use: Use an And node when you want entities that satisfy all of several conditions expressed as separate upstream branches. This is equivalent to combining multiple Criteria nodes with AND logic, but keeps the canvas visually clear.

Example: Two Criteria nodes — one for “high-value customers” and one for “opted-in to email” — connected to an And node to produce “high-value customers who are opted-in to email.”

Or

An Or node emits the union of all its upstream inputs — entities that appear in at least one connected upstream set (duplicates are removed).

When to use: Use an Or node when you want the combined pool from multiple conditions — for example, customers who triggered event A or who have attribute B.

Not

A Not node emits the complement of its single upstream input — entities from the canvas’s full entity model that do not appear in the upstream set.

When to use: Use a Not node to suppress a set — for example, connect an Audience Reference for “already-converted customers” to a Not node, then connect the Not node to an And node to exclude converters from your acquisition audience.

Shaping nodes

Shaping nodes partition, rank, or cap the set without changing its definition.

Split

A Split node partitions the incoming set into multiple labelled branches. Each branch becomes a separate path to a downstream Output node. Branches are disjoint by default — each entity appears in at most one branch.

Split modes:

When to use: Use a Split node for A/B testing (Percentage mode for stable assignment), multi-value-tier segmentation (Predicate mode), or randomised holdouts (Random mode).

Example (Percentage mode): A 70/30 split of a “lapsed customers” set into Treatment (70%) and Holdout (30%), each routed to its own Output node for separate activation.

Prioritize

A Prioritize node sorts the incoming set by a chosen attribute and retains the top N entities by rank.

When to use: Use a Prioritize node when you want to activate the highest-value or highest-risk entities — for example, “the top 5,000 customers by predicted lifetime value” or “the 1,000 most at-risk churners.”

Example: Prioritize on predicted_ltv, direction desc, top N 5000 → emits the 5,000 highest-predicted-LTV customers.

Limit

A Limit node caps the size of the incoming set by drawing an unbiased sample. For a top-N cut by a specific column, place a Prioritize node upstream of the Limit instead.

Limit modes:

When to use: Use a Limit node when you need to cap an audience size for budget, capacity, or experimentation reasons and want an unbiased sample — for example, a random 10,000-customer holdout for an A/B test.

Output node

Output

An Output node is a terminal node. Each Output produces a distinct, activatable audience. A canvas must have at least one Output node, and can have many.

When to use: Add an Output node at the end of every branch you want to activate. You can sync, use in journeys, or otherwise operate on each Output’s audience independently.

If you remove an Output node and the corresponding audience is still referenced by an active sync or journey, publishing is blocked. Resolve the conflict (remove the sync or journey reference first) before republishing.

Next Steps

• Building your first canvas — step-by-step walkthrough
• Publishing, versions & rollback — how publish, diff, and rollback work
• Canvas-owned audiences — how produced audiences behave