When OpenAI and Stripe announced the Agentic Commerce Protocol in late 2024, Walmart, Etsy, and Shopify launched integrations within weeks. Over a million merchants gained the ability to process purchases through ChatGPT almost overnight. The momentum has only accelerated since: Adobe data shows traffic to US retail sites from generative AI browsers and chat services increased 4,700% year-over-year in July 2025. 

McKinsey projects agentic commerce could generate up to $1 trillion in US B2C retail revenue by 2030, with global projections reaching $3 trillion to $5 trillion. For marketing and commerce leaders, the strategic question isn't whether this channel matters. It's whether you can measure it well enough to invest confidently.

Before scaling commitment to a new transaction channel, you need to know what visibility you'll have, what metrics you can capture, and how measurement capabilities compare to channels you already operate.

How agentic commerce actually works

The transaction flow is simpler than it might seem.

A customer opens ChatGPT and says "show me cordless drills under $200." The agent queries product feeds from integrated merchants, compares options based on the customer's criteria, and presents recommendations. The customer reviews the options and says "buy the second one." ChatGPT initiates a checkout session with the merchant through the Agentic Commerce Protocol, passing product selection, shipping address, and a secure payment token. The merchant validates the order, processes payment through Stripe, and fulfills it like any other transaction.

No website visit. No browsing session. No cart abandonment sequence. The customer expressed intent, the agent executed research and comparison, and a purchase happened through API calls rather than page loads.

This creates a measurement challenge that becomes clear when you map it against what marketers typically expect to see.

The metrics retail media networks depend on

Traditional e-commerce measurement gives marketing teams rich visibility across the customer journey. You track which campaigns drove traffic, how customers navigated your site, what products they viewed and compared, where they dropped off, and what finally converted. This data powers attribution, personalization, retargeting, and retail media network performance reporting.

The metrics that matter typically include:

Awareness and discovery metrics. Impressions, reach, and frequency for campaigns. Which placements drove initial product views. How customers first encountered your brand or products.

Consideration metrics. Product detail page views, time on page, comparison behavior. Add-to-cart rates, wishlist additions, save-for-later actions. These signals indicate purchase intent and inform retargeting.

Conversion metrics. Transaction completion, revenue, average order value. Conversion rate by traffic source, campaign, and audience segment. Return on ad spend for retail media placements.

Attribution metrics. Which touchpoints influenced the purchase. Multi-touch attribution across channels. Incrementality measurement proving that media spend caused conversions that wouldn't have happened otherwise.

Customer metrics. New versus returning customer identification. Lifetime value, purchase frequency, loyalty program engagement. Audience building for lookalike targeting and suppression.

These metrics serve specific purposes. Attribution proves media performance to advertisers. Consideration signals power retargeting and personalization. Customer identification enables loyalty programs and audience development. Retail media networks depend on this visibility to justify advertiser spend: Walmart alone generates $4.4 billion annually from retail media revenue.

The stakes for maintaining measurement capabilities are high. According to a recent EMARKETER and Bain survey, 48% of retail media network respondents say measurement and attribution issues are their top challenge, eMarketer and that's before agentic commerce adds a new layer of complexity.

What agentic commerce measurement looks like today

The Agentic Commerce Protocol was designed for transaction security, not analytics. It solves the hard problems of enabling agents to complete purchases safely: checkout session management, payment tokenization, fulfillment coordination. These capabilities made rapid merchant adoption possible.

What ACP doesn't include is measurement infrastructure. The protocol intentionally stays minimal.

Here's what merchants currently see from agentic transactions:

The consideration phase that drives so much measurement value in traditional e-commerce happens entirely in the agent's environment. By the time a checkout request arrives, you receive what you need to fulfill the order: product, address, payment. You don't receive what you need to understand the journey: what alternatives the customer considered, what factors drove the decision, whether any of your marketing influenced the purchase.

The data gap is substantial. Traditional e-commerce orders typically generate 40 or more data points: referral source, pages viewed, time on site, cart events, comparison behavior. Agent-mediated orders generate roughly six: order ID, items, total, timestamp, address, and payment method. Everything that informs marketing strategy is missing.

This creates specific gaps against the metrics that matter:

Awareness and discovery. No visibility. You don't know if the customer saw your ads, visited your site previously, or came to you purely through agent recommendation.

Consideration. No visibility. The comparison shopping happened in ChatGPT. You don't know what competitors the agent evaluated or why your product was selected.

Conversion. Partial visibility. You see the completed transaction but can't calculate conversion rate because you don't know how many consideration sessions didn't convert.

Attribution. No visibility. You can't connect the purchase to campaigns because referral data doesn't flow through agent transactions.

Customer identification. Partial visibility. You receive shipping and payment information but may not be able to connect this customer to previous purchases through other channels.

What measurement capabilities are emerging

The current state isn't the permanent state. Several developments are creating pathways toward better measurement.

Protocol extensions. OpenAI's roadmap references potential pre-checkout context that would allow agents to signal consideration events before purchase. If implemented, this could provide visibility into the discovery and comparison phases. Timeline and specifics remain uncertain.

Agent identification standards. Payment networks including Mastercard and Visa are developing "Know Your Agent" frameworks with cryptographic verification. These would help distinguish agent-mediated purchases from other API transactions and potentially enable agent-specific measurement. Google's Agent Payments Protocol uses signed mandates that prove what users authorized, providing verification infrastructure that could support measurement use cases.

Server-side capture at the trust boundary. The most actionable approach today involves capturing and enriching data at the point where requests first enter your infrastructure. This is the first legal and technical moment that user intent crosses into merchant control. Infrastructure at this boundary can identify the request source, append customer context from your existing systems, and create measurement records regardless of how the transaction originated.

Cross-channel identity resolution. When you can connect an agent-mediated purchase to a known customer profile, you recover significant measurement capability. The transaction itself provides identity signals (email, shipping address, payment method) that can link to existing customer records. This doesn't solve attribution for new customers, but it preserves customer metrics and lifetime value tracking for existing relationships. Research from Forrester found businesses with mature first-party data strategies achieve 2x increase in conversion rates and 30% reduction in customer acquisition costs, underscoring the value of strong identity infrastructure.

The metrics you can build toward

Even with current limitations, infrastructure investments now create measurement capabilities that grow as the ecosystem matures.

Transaction metrics by channel. Segment agent-mediated purchases from web, app, and in-store transactions. Track volume, revenue, and average order value by channel. Understand how the mix shifts over time.

Customer identification rate. Measure what percentage of agent transactions you can connect to known customer profiles. This indicates how much customer context you retain despite the measurement gap.

Repeat purchase behavior. Track whether customers who purchase through agents return through other channels or continue using agents. Understand cross-channel customer journeys even when you can't see the agent-side consideration phase.

Product performance in agent context. Identify which products convert through agent channels versus traditional channels. Differences may indicate how your catalog performs when agents evaluate specifications rather than customers browsing visually.

Time-to-fulfill and order quality. Compare fulfillment metrics between channels. Agent transactions may show different patterns in delivery preferences, return rates, or customer service contacts.

These metrics don't replace full-funnel measurement. They provide operational intelligence while infrastructure for consideration-phase visibility develops. As protocol extensions and identity standards mature, the foundation you build now extends to capture richer signals.

Building at the trust boundary

The architectural principle that enables agentic commerce measurement is straightforward: capture data at the point where requests enter your infrastructure, regardless of what initiated them.

This trust boundary is where your authority as data controller begins. Before this point, data exists in environments you don't control. After this point, you determine what happens. Infrastructure at this boundary creates measurement capabilities that work across channels.

Controller-side operation. Infrastructure running in your environment, under your control, following your rules. You own the data schema, enrichment logic, and routing decisions. When requirements change, you change the infrastructure.

Identity resolution at origin. Append customer context to inbound events while signals are fresh. Connect transactions to known profiles using the identity data you receive. Maintain consistent customer records across channels. Companies with proper CDP infrastructure show 2.9x greater year-over-year revenue growth versus those without, but CDPs can only work with what they receive. First-mile infrastructure ensures they receive complete, enriched data.

Consent verification before routing. Confirm what permissions apply to each transaction before data flows to downstream systems. Agent-mediated purchases may carry different consent context than web transactions. Enforce those differences at the collection layer.

Unified schema across surfaces. Normalize events from agents, web, mobile, and physical locations into the same structure. This creates a foundation for cross-channel analysis as measurement capabilities expand. With 59% of retail media networks citing enhanced measurement and reporting capabilities as their top strategic priority, infrastructure that provides consistent data across all transaction surfaces becomes a competitive requirement.

Companies like MetaRouter focus specifically on this first-mile infrastructure layer, operating at the trust boundary where controller-side capture, identity resolution, and routing happen before data reaches downstream systems. Whether you build internally or work with infrastructure partners, the architectural approach creates measurement capabilities that serve you regardless of how agentic commerce evolves.

Confidence before commitment

Agentic commerce measurement today is incomplete. The protocol layer solved transactions but didn't include analytics. Full-funnel visibility comparable to traditional e-commerce doesn't yet exist.

But incomplete isn't the same as impossible. Infrastructure at the trust boundary captures what's currently available and positions you to capture more as the ecosystem develops. Transaction metrics, customer identification, cross-channel behavior analysis, and product performance insights are buildable now. Attribution and consideration-phase visibility will follow as protocol extensions and identity standards mature.

The organizations approaching agentic commerce with confidence are building measurement infrastructure before committing to scale. They know what metrics they can capture today, what capabilities are emerging, and what foundation positions them to benefit as visibility improves.

The agents are already transacting. Building at the first mile with MetaRouter means you control measurement for this channel the same way you control it for every other channel. That visibility starts where your infrastructure begins.

‍