The evidence layer for agent-commerce.

AI agents and AI-mediated shopping surfaces can interact with merchant-controlled feed, storefront, cart, and checkout surfaces as part of shopping journeys. Those request paths create a measurement problem: merchants need to distinguish what was observed, what identity was declared, what could be verified, and what remains unknown.

Existing analytics, bot tools, feed tools, and commerce systems were designed for different jobs. Their records, taken separately, do not establish a shared merchant-side evidence chain for request identity, verification evidence, and observable commerce events across feed, storefront, cart, and checkout surfaces.

The planned Unidentified Traffic metric is the percentage of eligible storefront request events that remain unclassified after deterministic labeling. It is measurable only with complete server/CDN/origin log coverage. Cartograph is being built to provide evidence around that measure without inferring human identity or off-property model reasoning.

• Evidence over inference. Findings should separate what was directly observed from what was inferred, and label both, so teams know how much confidence to attach to them.
• Privacy by design. The planned observation layer is designed to avoid shopper PII, payment data, and unnecessary personal data. The focus is the request and the agent, not the human behind it.
• Merchant-side and observational. Evidence is designed to be captured from the merchant's own surfaces. Cartograph is not being built to make the purchase decision or execute the transaction.

As agent-mediated commerce develops, teams need a way to distinguish observed merchant-side evidence from declared identity, modeled output, and unknowns. That evidence layer is the operating foundation Cartograph is being built toward.