Demand source
Who introduced the journey
Preserve the first landing, trusted referral, and campaign context when available without treating them as checkout proof.
AI shopping journeys
Measure Shopify AI Shopping Agents Without Flattening the Journey
An AI referral, a later human visit, an agent-created cart, and an agent-executed checkout are different events. Shopify commerce evidence and explicit confidence should determine which claim the merchant can support.
Current product boundaryThe Shopify monitor operates in read-only OBSERVE during the private pilot. It does not process payments or hold funds.
THE OPERATING MODEL
Who introduced demand is not always who completed checkout.
An AI surface can introduce a shopper who later returns through search, email, direct traffic, or another channel. Conversely, an agent can execute a checkout after demand began elsewhere. A single source field cannot express every role in that journey.
AgentGateFee models demand source, last-touch source, checkout actor, checkout mechanism, execution confidence, and evidence separately. Missing evidence produces UNKNOWN or a lower-confidence classification rather than an invented agentic order.
WHAT THE RECORD MUST SHOW
Keep acquisition, assistance, and execution distinct.
Checkout actor
Who performed the transaction
Require commerce-side evidence before labeling an order agent-executed. A URL, app ID, or marketing Purchase event is insufficient on its own.
Confidence
How strongly the evidence supports it
Expose verified, supported, inferred, and unknown classifications with the evidence fields that produced each result.
EVIDENCE FLOW
A multi-touch evidence path
- 01Preserve the first attributable landing
- 02Record subsequent source transitions
- 03Read Shopify order attribution
- 04Identify the checkout mechanism when evidenced
- 05Assign execution confidence
- 06Compare economics by declared cohort
CLEAR PRODUCT BOUNDARY
A marketing event is not a commerce-side order fact.
AgentGateFee does not label an order AGENT_EXECUTED from a referral URL, app identifier, browser heuristic, or advertising event alone. The classification must be supported by evidence tied to the Shopify order or the executing commerce flow.
COMMON QUESTIONS
Understand the evidence before changing the workflow.
Is every AI-referred order agent-executed?
No. AI-referred demand and agent-executed checkout describe different roles and can occur in different sessions or channels.
Can a journey have more than one source?
Yes. The model can preserve first demand source and later touchpoints separately instead of forcing one label to explain the entire journey.
What happens when execution cannot be proven?
The result remains UNKNOWN or a lower-confidence classification, and the missing evidence stays visible.
START WITH EVIDENCE