The 393 Percent Channel Nobody Is Measuring
In the first quarter of 2026, commerce telemetry from Triple Whale and Adobe Analytics recorded something striking. Across the ecommerce queries they track, AI assistants generated roughly 606,000 product citations. That is a 393 percent year-over-year increase. It is faster growth than any other acquisition channel in retail. Moreover, AI shoppers were higher intent. Specifically, they converted at 3.0 percent versus a 1.4 percent baseline for traditional ecommerce traffic. In several verticals, the gap was wider still. For example, hotels and resorts converted AI-referred traffic at roughly 7.0 percent. Likewise, legal services hit 5.6 percent. Similarly, healthcare hit 4.5 percent.
Bonus
Download a PDF version of this blog. Access it offline anytime. Bring it to team or client meetings.
The attribution gap
However, that headline number hides a more uncomfortable one. Specifically, around 70 percent of AI-driven referrals are misclassified as direct traffic in default GA4 setups. The figure comes from a May 2026 aggregation by Elogic Commerce. Notably, Elogic compiled data from 28 primary sources, including Adobe Analytics, Triple Whale, Similarweb, and tier-1 analysts.
There are three mechanisms at work. First, paid ChatGPT accounts strip referrer data. Second, Gemini’s Deep Research mode does the same. Third, Perplexity’s outbound clicks frequently arrive unattributed. As a result, AI-referred traffic is undercounted by an estimated factor of three to four. Consequently, a generation of ecommerce operators sees a quiet uptick in direct traffic. Meanwhile, they are missing the most consequential channel shift since mobile commerce.
The shift is no longer hypothetical
Importantly, the shift is now happening in production. In September 2025, OpenAI launched Instant Checkout inside ChatGPT. Launch partners included Etsy, Shopify, Walmart, Target, Sephora, Nordstrom, Best Buy, Lowe’s, and Home Depot. Subsequently, Google released its Universal Commerce Protocol in January 2026. Furthermore, Google rolled out agentic checkout inside Gemini and AI Mode in Search. Meanwhile, Shopify is rolling out Agentic Storefronts through 2026.
Notably, the stakes are now also being litigated. On November 4, 2025, Amazon sued Perplexity in the Northern District of California. The complaint runs under the federal Computer Fraud and Abuse Act. Specifically, Amazon wanted to keep AI agents off its platform. In March 2026, a federal judge granted Amazon a preliminary injunction. As a result, the case frames just how high the stakes are.
You Might Also Like
The analyst data points the same direction. First, eMarketer puts AI-driven retail spend in 2026 at $20.9 billion. That is roughly quadruple 2025’s figure. Second, McKinsey projects $3 to $5 trillion in global retail spend will be redirected through agentic channels by 2030. Third, Bain (via MetaRouter) projects 15 to 25 percent of total online retail will flow through agents by the end of the decade.
Why this is an architecture story, not a marketing one
Importantly, none of this happens without an architecture shift on the merchant side. Specifically, AI agents do not browse storefronts the way humans do. They do not see Liquid themes. They do not appreciate beautiful product photography. Likewise, they do not navigate marketing funnels. Instead, they parse structured data. They query APIs. They execute transactions through open protocols.
Therefore, the merchants who become agent-ready in 2026 will capture the curve. Conversely, the merchants who do not will become invisible to a generation of buyers. Notably, this is not because their products are inferior. Rather, it is because their storefronts cannot be read.
This post is the operating brief for ecommerce leaders making 2026 architecture decisions. Specifically, it synthesizes the freshest data: Deloitte’s 2026 Retail Outlook, McKinsey’s October 2025 agentic commerce analysis, the OpenAI ACP and Google UCP protocol launches, and Shopify’s Agentic Storefronts rollout. Ultimately, the thesis is simple. The agentic commerce shift is not a marketing problem. Instead, it is an architecture problem. And the architecture problem has a name. It is called headless.
What Agentic Commerce Actually Is
First, the definition. Essentially, agentic commerce is a shopping flow in which an AI agent acts on behalf of the buyer. Specifically, the agent does four things. First, it discovers products across merchants. Second, it compares attributes against the buyer’s stated constraints. Third, it negotiates where possible. Fourth, it completes the transaction. Notably, OpenAI’s own framing captures the shift cleanly. ChatGPT doesn’t just help you find what to buy. It also helps you buy it.
Three things make agentic commerce categorically different from earlier AI-assisted shopping. First, the agent now has transactional authority. Specifically, it can complete the purchase, not just produce a comparison table. Second, the agent operates across merchants. As a result, discovery and ranking happen outside the merchant’s controlled environment. Third, the agent transacts via open protocols. Consequently, the merchant either implements the protocol or is excluded from the transaction. Ultimately, the shift is from a recommendation surface to a transaction surface. That difference is exactly the difference between marketing and commerce.
The three transaction modes live in the market right now
As of mid-2026, three transaction mechanics are operating in production. Therefore, understanding which one applies to your merchant model is the first practical step.
First, protocol-based transactions are the cleanest of the three. Specifically, they run over open standards like OpenAI’s ACP and Google’s UCP. In this mode, the agent negotiates pricing, inventory, and checkout fields directly with the merchant’s commerce APIs. Then, the transaction completes inside the agent surface. Importantly, the buyer never visits the merchant’s storefront. For example, this is the model used by Shopify merchants integrated with OpenAI’s Instant Checkout.
Second, redirect transactions are the practical fallback. In this mode, the agent does the research and recommendation work. Then, it hands the buyer a deep link to the merchant’s product page or cart. Finally, the buyer completes checkout on the merchant’s own site. Consequently, this is what most agentic shopping looked like at the end of 2025. Today, it remains the default for merchants without protocol integration. Notably, conversion rates are still elevated relative to traditional traffic. However, the merchant loses the seamless experience of fully protocol-based flows.
Third, hybrid app experiences are the most strategically ambiguous mode. Here, the merchant builds a custom integration. Examples include a Shopify Agentic Storefront, a Salesforce Commerce Cloud agent surface, or a bespoke build on a composable stack. Specifically, the merchant’s product catalog appears directly inside the agent’s interface. In short: most controlled, most architecturally demanding.
The Numbers Behind the 2026 Shift
Here is the consolidated 2026 picture from the analyst and platform data published in the past six months:
| Metric | 2026 Value | Source |
| AI-driven retail spend (2026) | $20.9 billion | eMarketer (Dec 2025) |
| Global retail spend redirected by 2030 | $3T – $5T | McKinsey (Oct 2025) |
| Share of online retail via agents by 2030 | 15% – 25% | Bain via MetaRouter |
| Retailers planning to adopt agentic AI | 68% in next 12–14 mo | Deloitte 2026 Retail Outlook |
| ChatGPT eCommerce conversion rate | 3.0% (vs 1.4% baseline) | Elogic Commerce (May 2026 aggregation) |
| AI-channel traffic YoY growth (Q1 2026) | +393% | Triple Whale / Adobe Q1 2026 |
| AI citations across eCom queries (Q1 2026) | ~606,000 | Triple Whale Q1 2026 |
| AI referrals misclassified as ‘direct’ in GA4 | ~70% (estimate) | Elogic Commerce 2026 |
| Agentic AI market by 2032 | $93.2B projected | Multiple analyst syntheses |
The data tells two stories simultaneously. On one hand, AI commerce is still a small slice of total ecommerce traffic. Specifically, it is well under one percent today by most measures. Given that base rate, the temptation is to defer the architecture work. On the other hand, this is the fastest-growing acquisition channel in retail. Furthermore, the conversion rates are categorically higher than traditional channels. Notably, the entrants are the largest technology platforms in the world. The base rate is small. The trajectory is not.
Which verticals benefit first
Importantly, the vertical conversion data tells you who the early winners will be. Specifically, AI agents do their best work in two kinds of categories. First, categories where the buyer arrives with constraints that are hard to compare manually. Second, categories where the purchase carries enough intent to justify a research-and-execute flow.
| Vertical | AI Conversion | Why It Outperforms |
| Hotels and resorts | 7.0% | High-research, high-intent purchase; AI excels at multi-criteria filtering. |
| Legal services | 5.6% | Buyers arrive with a specific need pre-articulated by the agent. |
| Healthcare | 4.5% | Information-dense category where AI synthesis adds genuine value. |
| General ecommerce (avg) | 3.0% | ChatGPT-referred shoppers across categories. |
| Traditional ecommerce traffic baseline | 1.4% | Cross-channel ecommerce average. |
Source: Elogic Commerce 2026 aggregation; figures reflect ChatGPT-referred conversion rates across the verticals where vendor data was available. Vertical performance varies considerably by sub-category.
The implication for category strategy
Notably, high-research, high-intent categories see the biggest premium. Examples include travel, B2B SaaS, professional services, healthcare adjacencies, and considered DTC. Specifically, the agentic-commerce conversion premium is two-to-five times the cross-channel baseline. Accordingly, the case for protocol readiness in 2026 is materially stronger for these categories. Conversely, low-consideration commodity categories see less marginal value over established channels.
Why Server-Rendered Storefronts Will Lose (Liquid Included)
This is the architecture argument. Notably, it is the part of the analysis that ecommerce executives most often resist. Specifically, the instinct is to treat agentic commerce as a marketing or SEO problem. For example, executives reach for generative engine optimization (GEO), structured data plugins, and schema markup. However, that instinct is only half right. GEO work is necessary. It is also nowhere near sufficient.
What agents actually see
Importantly, AI agents do not see what shoppers see. Specifically, they do not parse rendered HTML. This is true for Liquid themes, server-rendered Magento pages, and server-rendered WooCommerce templates. Instead, they request structured data over APIs.
Consequently, the flow that follows a ChatGPT recommendation depends entirely on what the storefront exposes. Specifically, the agent needs five things. First, structured product metadata. Second, real-time inventory. Third, authoritative pricing. Fourth, programmatic cart and checkout endpoints. Fifth, compliance with at least one agentic commerce protocol.
Notably, a traditional storefront fails on all five. For example, product specs sit inside HTML descriptions. Likewise, the add-to-cart flow requires human form-field input. Furthermore, the inventory feed is not agent-readable. Finally, there is no programmatic checkout pathway. In short, this is not a slow agentic experience. It is an invisible one.
The three technical gates
Weaverse, a Shopify Hydrogen build studio, framed the constraint clearly in March 2026. Specifically, AI agents parse metafields to understand products. Consequently, if product specs live only in HTML descriptions, AI cannot read them.
Furthermore, the same logic applies one layer deeper. Specifically, AI agents need three things. First, consistent JSON-LD schema. Second, fast time-to-first-byte. The industry target sits between roughly 200 and 400ms depending on the source. Third, predictable URL patterns.
Importantly, server-rendered Liquid themes miss on all three counts. Notably, they were optimized for human browsers and SEO crawlers from a previous era. Conversely, Hydrogen storefronts meet the criteria natively. Specifically, Hydrogen uses a headless commerce architecture. The front end runs on React. The back end runs on Shopify Storefront APIs.
The mental model: agents are headless customers
Imagine your largest customer segment cannot see your storefront. Specifically, they cannot read your product photography. They cannot navigate your nav menu. Likewise, they cannot fill out your checkout form by hand. Therefore, every piece of information they need must arrive as structured data through an API. Furthermore, every action they take must be programmatic. If your architecture cannot serve that customer, then the agents acting for that customer will serve a competitor instead.
The Open Protocols That Decide Who Sells
The most consequential shift of 2026 is not the volume of AI shoppers. Instead, it is the formalization of agentic commerce protocols. Until 2025, every AI shopping integration was a one-off arrangement. Specifically, deals ran between a platform (OpenAI, Google, Microsoft) and a merchant. However, that model does not scale. Therefore, through 2025 and into 2026, the platforms have published open protocols. These protocols make agent-to-merchant transactions interoperable.
| Protocol | Vendor / Origin | Launched | What It Does |
| ACP (Agentic Commerce Protocol) | OpenAI + Stripe | Sept 2025 | Lets ChatGPT initiate transactions with merchants. |
| UCP (Universal Commerce Protocol) | Jan 2026 | Gemini + Search ‘Buy for me’ actions. | |
| MCP (Model Context Protocol) | Anthropic (open) | Nov 2024+ | Open standard for agent ↔ tool integration. |
| Shopify Agentic Storefronts | Shopify | Rolling out 2026 | Native agent surface for Shopify merchants. |
The strategic implication is straightforward. Specifically, protocol support becomes a gate. For example, merchants who implement ACP can complete transactions inside ChatGPT. Conversely, merchants who do not, cannot. They may still appear in product recommendations. However, the buy button is grayed out. Alternatively, the agent redirects the buyer to a competitor that does support the protocol. Notably, the same logic applies to Google UCP for Gemini and AI Mode in Search. Likewise, it applies to Shopify Agentic Storefronts. Ultimately, the agentic ecosystem is forming around a handful of protocols. Consequently, protocol membership decides which merchants can participate in which transactions.
The Agent-Ready Headless Architecture
Translating the agent-readiness requirements into a concrete pattern produces a recognizable shape. Specifically, a headless or composable commerce stack with five clearly separated layers. Furthermore, an API gateway designed to serve both humans and agents. Finally, a content-and-product-data layer built for structured retrieval rather than rendered display.
Notably, this is not a radical pattern. Specifically, it is the MACH architecture that the composable commerce movement has been advocating for half a decade. MACH stands for microservices, API-first, cloud-native, and headless. Importantly, it is now finally meeting a buyer behaviour that demands exactly its strengths.
Layer 1: The consumer layer is now multi-persona
First, the consumer layer is no longer a single persona. Instead, it is at least three. First, human browsers (web, mobile, app). Second, AI agents (ChatGPT, Gemini, Copilot, Perplexity). Third, other channels (voice assistants, social commerce, marketplaces, point-of-sale). Notably, each persona makes different requests against the same commerce backend. Therefore, the architecture must give no persona privileged access to the data.
Layer 2: The protocol layer is the new 2026 addition
Second, the protocol layer is the genuinely new addition in 2026. Specifically, ACP, UCP, MCP, and Shopify’s Agentic Storefronts framework all introduce expectations the merchant must meet. Examples include structured request/response formats, authentication patterns for agent-initiated transactions, idempotency guarantees, and dispute-resolution semantics for agent-completed orders. Although merchants do not need to support every protocol from day one, the architecture must be able to add new protocols later. Notably, this must happen without rewriting the commerce backend. Therefore, protocol adapters live in this layer. Commerce logic does not.
Layer 3: The experience API layer rebuilt for agents
Third, the experience API layer is where agent requests get adapted into the commerce backend’s native query language. Typically, it is built as a backend-for-frontend (BFF) or a GraphQL federation gateway. Specifically, three things happen here. First, structured JSON-LD responses get assembled for agents. Second, rate limiting protects the backend from agent traffic spikes. Third, authentication for agent-initiated transactions gets validated. Consequently, a BFF designed only to serve a React storefront will not survive contact with agentic traffic. Therefore, the layer needs to be rebuilt with agents as a first-class consumer.
Layer 4: Composable commerce services
Fourth, the commerce services layer is where composable commerce pays off. Specifically, the layer covers product, cart, order, payment, fulfillment, and subscription. Furthermore, best-of-breed services can be wired together behind the API gateway. Examples include commercetools, Shopify, BigCommerce, Salesforce Commerce Cloud, and open-source options like Saleor and Medusa. Critically, every service in this layer must be programmatically addressable. For example, cart and checkout services that assume a human-driven UI flow break in agentic contexts. Conversely, cart and checkout services with clean APIs do not.
Layer 5: Content and data foundation
Finally, the content and data layer is where most traditional merchants are weakest. Specifically, a headless CMS solves the problem. Examples include Contentful, Sanity, Strapi, Storyblok, or equivalents. Furthermore, pair the headless CMS with a proper product information management (PIM) system. Together, these give the merchant a structured authoring environment. As a result, product specs, marketing copy, and brand content are all stored as queryable data. Importantly, this is the layer that makes the merchant readable to AI agents at all.
The trust, fraud, and dispute-resolution problem
Notably, the architecture conversation gets the most coverage. However, the operational risk conversation is the part that keeps merchants from saying yes. Specifically, the Amazon vs. Perplexity case is the canonical example.
On November 4, 2025, Amazon filed a federal complaint against Perplexity. The complaint accused Perplexity of using its Comet browser to disguise agent traffic as regular Chrome sessions. Furthermore, the complaint said Comet logged into Amazon accounts and completed purchases without authorization. Subsequently, in March 2026, a federal judge in San Francisco granted Amazon a preliminary injunction. As a result, Comet is now blocked from buying on Amazon while the case proceeds. Importantly, the underlying questions remain unresolved across the industry. Therefore, merchants implementing agent-ready architectures need to design around them deliberately.
Four operational questions to answer before rollout
Concretely, four operational questions deserve explicit answers before any production rollout:
- Agent authentication and provenance.Specifically, can the merchant cryptographically verify the order? Notably, the order must come from an authenticated agent acting on behalf of a known buyer. Otherwise, it could be a scraping bot or an account-takeover attempt. For example, OpenAI’s ACP and Google’s UCP both define agent-identity primitives. Use them.
- Dispute and chargeback workflows.Furthermore, what happens when the agent buys the wrong item, size, or price? Specifically, whose customer service queue handles the return? Likewise, whose payment instrument processes the refund? Importantly, merchants currently shipping protocol-based flows document these answers in their integration playbooks. Conversely, they do not assume the platform will handle it.
- Data controller responsibilities.Additionally, when an agent acts on a buyer’s behalf, which party is the data controller? Specifically, the merchant, the agent platform, or the buyer? Notably, different jurisdictions answer this differently. As a result, the answer changes the compliance footprint.
- Fraud signal recalibration.Moreover, agent traffic looks different from human traffic and different from bot traffic. Consequently, fraud-detection systems tuned on human signal patterns will produce false positives on agent traffic. Likewise, they will produce false negatives on agent-disguised fraud. Therefore, plan for a fraud-system tuning phase as part of every protocol rollout.
Importantly, the merchants who treat these questions as architectural inputs will avoid the operational surprises. Conversely, those who treat them as legal afterthoughts will face surprises like the ones now playing out in Amazon’s courtroom.
Practical Takeaways: What to Do This Quarter
For ecommerce leaders making 2026 plans, here is the prioritized action list. Importantly, none of these require completing the agent-ready transition this quarter. However, all of them require starting it this quarter.
Foundation: audit, measure, identify
- First, audit your structured product data.Specifically, walk through your top 50 SKUs. Then, answer one question: are all the specs stored as structured metadata? Examples include size, weight, materials, compatibility, and certifications. Alternatively, are they buried inside HTML description fields? If the answer is HTML, then you have weeks of PIM cleanup ahead. Notably, this work must precede any agent integration.
- Second, measure your true AI-channel traffic.Furthermore, set up UTM-based AI referral tagging across the major AI platforms. Examples include ChatGPT, Gemini, Copilot, and Perplexity. Then, reconcile against your direct traffic line. Assume your current AI traffic is undercounted by 3–4×. Without this baseline, no 2026 investment can be justified.
- Third, identify your protocol gates.Additionally, list the AI surfaces that matter most to your category. For example, ChatGPT for general retail. Likewise, Gemini for Google-search-driven categories. Similarly, Perplexity for high-research verticals. Then, check which protocols each requires. Notably, ACP, UCP, and Shopify Agentic Storefronts each unlock different surfaces.
- Fourth, evaluate your storefront architecture honestly.Moreover, if you are on a server-rendered platform, document the agent-readiness gaps. Examples include Liquid themes, Magento monoliths, and server-rendered WooCommerce. Specifically, audit structured data, TTFB, and programmatic checkout. Consequently, the gap analysis becomes the input to your 2026 replatforming budget conversation.
Execution: migrate, pilot, attribute, invest
- Fifth, plan a phased headless migration if needed.Importantly, big-bang replatforming has a poor record in commerce. Conversely, strangler-pattern migrations have a much better one. Specifically, replace the front end first. Then, progressively decouple commerce services behind a BFF. Start with the storefront surfaces that matter most to AI agents. Examples include PDPs, category pages, and search. Meanwhile, leave back-office systems untouched.
- Sixth, pilot one agentic integration end-to-end.Furthermore, pick the protocol that matches your highest-priority AI surface. For example, ACP for ChatGPT. Similarly, UCP for Gemini. Likewise, Shopify Agentic Storefronts if you’re on Shopify. Then, integrate one product category through the full transaction flow. Notably, the operational learnings are not available in any whitepaper. They come from the integration. Specifically, you learn about dispute resolution, refund handling, attribution, and customer service for agent-initiated orders.
- Seventh, rebuild your attribution model.Additionally, if roughly 70 percent of AI referrals land in your direct-traffic bucket, then your acquisition cost calculations are systematically wrong. Therefore, work with your analytics team to layer AI-aware attribution on top of standard GA4. Examples include server-side tagging, referrer reconstruction from HTTP headers, and dedicated UTM conventions for agent traffic.
- Eighth, invest in agent-readable content.Finally, structured FAQs, comparison tables, technical specs, and review summaries are exactly the content that AI agents query. Consequently, the merchants who appear in AI recommendations have the most agent-readable content libraries. Notably, this is content marketing for a non-human audience. Furthermore, it is a real 2026 budget line.
The Strategic Window: Why 2026 Matters More Than 2027
Three structural reasons make 2026 the year the architecture decision matters most. Notably, none of them apply equally to 2027.
First, the platforms are still negotiating which protocols dominate. Specifically, OpenAI’s ACP, Google’s UCP, Anthropic’s MCP, and Shopify’s Agentic Storefronts framework are all live. Furthermore, they all overlap in scope. Importantly, they are all collecting design feedback from merchants. Consequently, the merchants who integrate early get input into how the protocols evolve. Conversely, the merchants who wait integrate into a finished standard. As a result, they will have no influence on its direction.
Second, the competitive base rate is still low. Specifically, AI traffic sits at well under one percent of total ecommerce volume. Therefore, the cost of being agent-unready in 2026 is missed upside. It is not yet lost market share. However, by the time AI traffic hits 5 to 10 percent, the cost flips. Notably, Bain projects this for the late 2020s. Then, agent-unready merchants will visibly lose share to agent-ready competitors. Furthermore, the architecture investment that costs $1 million in 2026 will cost $5 million in 2028. Specifically, it will cost more under crisis pressure with fewer qualified engineers available.
A note on platform economics
Third, the rails are currently free. Importantly, they will not stay that way. Specifically, OpenAI, Google, and Shopify are introducing protocol-based commerce as an organic acquisition channel. However, every major platform that began as organic distribution eventually monetized. Examples include search, social, marketplaces, and app stores. Likewise, each monetized through take-rates, sponsored placement, or ad inventory.
Therefore, merchants who establish protocol presence while the rails are organic gain a real advantage. Specifically, they will negotiate the eventual take-rate from a position of strength. Conversely, those who wait for the rails to settle will be onboarding into a paid channel. Furthermore, they will be onboarding against incumbents who already know how to operate it.
For PracticalLogix’s commerce customers, the framing is this. Specifically, the agentic commerce transition is not the next ecommerce trend. Instead, it is the next ecommerce platform shift. Notably, it is on the same scale as the desktop-to-mobile transition of the early 2010s. Importantly, merchants that recognized the mobile shift early captured a decade of share gains. Conversely, merchants that waited spent the rest of the decade catching up. Today, the same dynamic is forming around agentic commerce. Ultimately, the architecture decisions made in the next two quarters compound for the rest of the decade.
Conclusion: From a Marketing Problem to an Architecture Problem
Most ecommerce leadership teams arrived at agentic commerce through their marketing function. Specifically, their first instinct was to ask the SEO team how to rank in AI search results. That instinct was reasonable in 2024. However, it is incomplete in 2026. Notably, ranking in AI recommendations is necessary but insufficient. It gets the merchant considered. Conversely, being purchased requires architecture. Specifically, the merchant needs four things. First, structured data the agent can read. Second, APIs the agent can call. Third, protocols the agent can transact through. Fourth, a programmatic checkout path the agent can complete without human intervention.
Who wins and who loses
Consequently, the merchants who solve the architecture problem will dominate the next wave of ecommerce growth. Conversely, the merchants who treat it as a marketing problem will struggle. Specifically, they will spend the rest of the decade explaining to their boards why the 393 percent channel passed them by.
Notably, none of this is a prediction about a distant future. Today, ChatGPT is buying products on behalf of real shoppers with real money. Specifically, it runs on the merchant integrations OpenAI signed before September 2025. Similarly, Gemini is doing the same through Google’s January 2026 protocol launch. Meanwhile, Shopify is rolling out Agentic Storefronts to merchants through 2026. Importantly, the infrastructure for agentic commerce is shipping in production every quarter.
The price of admission
Therefore, the only open question is which merchants invest now. Specifically, which merchants build the architecture that lets them participate. For PracticalLogix, the question we bring into every 2026 planning conversation is simple. Is the storefront built for one consumer persona or three? Ultimately, headless commerce, structured product data, and protocol-ready checkout are no longer architectural preferences. Instead, they are the price of admission. They define the channel that will define the next decade of ecommerce.
