The basic layers of selling online have largely been solved. The new contest, a growing number of commerce leaders argue, is happening inside the transaction itself. Here is what the modern commerce stack actually looks like, according to Rokt, commercetools, Shopify, and other commerce leaders, and why the purchase moment has become its most valuable real estate.
For almost the entire last decade, e-commerce businesses were built by assembling a familiar set of tools. These include a storefront, a catalog, a checkout, a payments processor, a fulfillment system, an analytics suite, and a lifecycle marketing platform. All you had to do was get the pieces working together, drive traffic to the funnel, and process orders efficiently. This is what worked and has been working until now.
It also produced a problem. As those basic layers matured, they did stop being a source of competitive advantage. Marketplace, checkout, payments, and fulfillment increasingly look the same from one company to the next. So a question now sits in front of commerce leaders planning their next phase of investment: if the basics are commoditized, where does the next layer of value come from?
A growing body of thinking points to an answer that is less about adding new systems and more about coordinating the ones companies already own, with a focus on a single high-value window. In a detailed guide for commerce leaders, the New York-based e-commerce technology company Rokt makes the case that the frontier is no longer the funnel leading up to checkout or the campaigns that follow it. It is the transaction itself, the window from product selection through confirmation, when purchase intent is highest and behavioral signals are live.
What the modern commerce stack actually is
If we had to explain in detail, the modern commerce stack powers how a business sells online, spanning traffic acquisition, product discovery, cart and checkout, payments, fulfillment, analytics, and post-purchase retention. Nothing in that list is new. But the addition of an emerging layer designed to act inside the purchase flow rather than around it is what makes it worth considering.
mCommerce leaders describe the foundation of that stack in two different ways. Platform companies such as Shopify frame it as unified commerce, a single system that runs sales online, in-store, and everywhere in between for millions of businesses across 175 or more countries. Composable specialists such as commercetools consider it as a best-of-breed assembly of independent, API-first components that a business can swap and scale on its own terms. Both camps now arrive at the same conclusion: once the foundation is in place, the next contested tier is the transaction itself.
Checkout doesn’t have to be this dead zone between getting a customer and keeping one. Three things work together here to figure out what a customer actually needs to see mid-purchase. Sometimes nothing. We’re not stuffing more messages into checkout — that’s not the point. The point is speed: decisions happen fast enough to match how someone actually shops, so it doesn’t feel bolted on. Feels like part of the purchase. And you can measure it, unlike most checkout noise.
● Real-time data: capturing and activating first-party signals as they happen, rather than syncing them in batches after the fact.
● Decisioning systems: determining the next best action from live session inputs such as cart contents, basket value, and product category.
● In-purchase experiences: delivering the resulting interaction inside the transaction flow without disrupting checkout.
The defining characteristic, Rokt’s guide stresses, is low latency between those three. These capabilities are meant to increase the value of each transaction, not to replace the transaction itself.
Why the purchase window became the prize
If we had to reason this out, we’d say the reasoning rests on a point about signals. Pre-purchase marketing usually leans on historical profiles, predicted intent, and audience segments. But post-purchase marketing will react only after the sale is done. The window in between, what Rokt has trademarked as the Transaction Moment, is the one stretch of the journey where a customer is actively completing a purchase. That way, a business can still use live context to shape what happens next.
Everyone’s calling it commerce media now. IAB‘s number: $63.4 billion in 2025, up about 18% from last year. That’s faster growth than almost anything else in digital advertising. The Silicon Review had a good line on this recently, pulling from Rokt’s research, that the moment between picking a product and paying for it is basically where commerce lives or dies now. Not before. Not after. Right there.
Rokt’s chief commercial officer Elizabeth Buchanan didn’t mince words about it in the company’s 2026 outlook. Commerce media has hit an inflection point, she argued, and scale without control just leads to saturation and consumers tuning out. The winners won’t be whoever shows the most banners. They’ll be whoever owns the highest-value moment in the customer’s journey.
Where traditional stacks fall short
Most companies at this stage aren’t short on tools. They’ve usually got five or six stitched together already. The problem is coordination between them. So data lands in a CDP or CRM through a batch sync. By the time it’s processed, the transaction’s long finished. Decisioning is working off segments built weeks or months ago instead of anything happening in the actual session in front of it. And the experiences themselves? Locked in ahead of time. Campaigns planned out, post-purchase emails scheduled, none of it reacting to the moment it actually shows up in front of someone.
Latency in these setups usually runs minutes to hours. Most of what counts as optimization is really just friction reduction, which isn’t nothing, but it’s not the same thing. By the time anything relevant reaches a customer, they’ve already bought (or already left).
Flip all of that, and you get a modern stack. No more waiting on a batch sync because data will stream in as it happens. In this scenario, segments from last month don’t drive the decision anymore. Instead, you need to focus on what’s happening in the session right now. Even the experience changes: instead of a campaign somebody scheduled two weeks ago, it lives inside the purchase flow itself, reacting in real time instead of running off a script.
Milliseconds, sometimes less- that’s the latency now. And it’s not really about the conversion anymore either; it’s about squeezing more value out of each one once it’s happening.
Usually the components haven’t changed much. It’s the wiring between them, and the speed, that’s different.
The four layers and the tools that fill them, page 3
The guide splits a high-performing stack into four layers, and it doesn’t shy away from naming names, competitors included, not just Rokt’s own tools. That’s sort of the point it’s making: no one vendor owns this whole thing. The value’s in how the layers actually connect.
● Data layer: captures transactional and behavioral signals in real time. Common tools include Segment, Snowflake, RudderStack, and Rokt mParticle.
● Decisioning layer: determines the next best action from current session data. Common tools include AWS SageMaker, in-house ranking models, Bloomreach, Dynamic Yield, and Rokt.
● Experience layer: delivers the interaction within the purchase flow. Common tools include Rebuy, Bold Commerce, Stripe, and Rokt.
● Value layer: measures impact on incremental revenue, average order value, and conversion. Common tools include Amplitude, Looker, Northbeam, and Triple Whale.
Sub-100 milliseconds. That’s the bar Rokt uses for the surfaces where speed actually matters. Surprisingly, it’s not a nice-to-have. As soon as you blow past that window, people actually get to feel it, even if they couldn’t tell you why the checkout felt off. Hit it, and there’s nothing to notice. It just works.
A natural fit with the composable shift
This layered approach maps onto a wider architectural move that wasn’t dominating for years. Rather than buying a single monolithic platform, more retailers now assemble systems from specialized, API-first components. Let’s consider this as an approach the industry calls composable commerce. The large majority of enterprise companies have considered or moved toward composable architecture, and the event-driven design that underpins it is precisely what allows live signals to flow between services fast enough to act on.
Few companies have argued that case longer than commercetools, which helped popularize composable commerce and the MACH approach (microservices, API-first, cloud-native, and headless) that sits beneath it. The company identifies three traits it says separate genuine composability from marketing claims: cloud-native SaaS, components that are independent both technically and commercially, and a technology-agnostic architecture. Michael Scholz, a commercetools product and marketing leader, describes composable as decoupling the entire stack and choosing a best-of-breed module for each function, so any piece that stops serving the business can be replaced without disturbing the rest. That independence is what lets a real-time decisioning or in-purchase experience layer plug in without a full replatform. The modern commerce stack, in other words, is the application of composable thinking to the one moment where speed matters most.
The unified counterpoint, and why both roads lead to the transaction
Not every leader buys the best-of-breed thesis, and the most-watched counterargument comes from the platform side. Shopify, which says it powers commerce for millions of merchants from entrepreneurs to brands such as SKIMS, Vuori, and Supreme, makes the case for unified commerce: a single platform that reduces the custom code and operational burden a fully assembled stack can create. Shopify argues that architecture decisions have an outsized effect on transformation speed. The difference between a nine-month replatform and a roughly 90-day launch often comes down to whether teams choose systems that require heavy customization or a composable approach built for flexibility out of the box. The company points to results such as Skullcandy, which it says simplified its stack and posted 45 percent year-over-year revenue growth.
Most likely, it turns out the distinction matters less than you’d think. Consolidated platform or composable stack, it doesn’t really matter. Instead, the basic layers are table stakes either way now. What’s actually contested is the layer sitting on top. Rokt, commercetools, and Shopify all came at this from different directions and landed on the same idea: the transaction moment is where you win or lose now, and whatever your stack looks like, it needs to be fast enough and coordinated enough to act right there.
Craig Galvin, Rokt’s chief revenue officer, put it well. According to him, the transaction moment might be the most valuable, most underused interaction in commerce. Intent’s clear. Attention’s still there. That combination doesn’t last.
What it looks like in practice
If you consider having a real-world example, Rokt’s probably the clearest one available right now. Even analysts and outside coverage keep pointing to it as a reference case. Gartner named the company in its Market Guide for Retail and Commerce Media Networks, way back in April 2026. It was called out as an example of the emerging post-purchase category. Even Yahoo Finance and MarTech Cube covered it.
You can see the mechanics of this play out in a partnership that got a lot of attention. Fanatics returned to the Rokt Network and started rolling out two products right in that experience layer, one that works on the payment page itself, surfacing offers in real time, and one that kicks in right after someone buys. Trade press covering it said the rollout covers Fanatics.com plus team and league sites across North America, Europe, and Asia-Pacific.
Cineplex did something similar back in January 2026, folding Rokt’s tech into checkout so moviegoers see relevant offers as they’re buying tickets.
A detail from that coverage captures what separates this layer from conventional advertising. The system is built to show nothing at all when a relevance threshold cannot be met. Claire Southey, Rokt’s chief AI officer, summarized the philosophy by saying the best move is often to get out of the customer’s way. Instead, you must let them complete the purchase. Restraint, in this model, is a performance feature rather than a missed opportunity.
The scale behind the thesis
The argument carries more weight because of the volume behind it, across all three companies. Rokt’s network is on track to power more than 10 billion transactions in 2026 across thousands of leading e-commerce businesses, with revenue growing to more than 800 million dollars in 2025. That growth has not escaped notice by the wider industry: the company placed 243rd on the 2025 Deloitte Technology Fast 500, with revenue more than tripling over the three-year measurement period. The same network now serves companies including Live Nation, Macy’s, PayPal, AMC Theatres, and Uber.
Underneath all this, the platform layer’s hitting its own scale records. Shopify said merchants on its platform cleared more than $100 billion in gross merchandise volume in just the first quarter of 2026. Harley Finkelstein, the company’s president, has been calling this the AI era of commerce, with Shopify positioned as a “category of one,” pointing to investments in catalog, AI assistance, and a universal commerce protocol.
Meanwhile, composable adoption keeps spreading through the enterprise tier. Commercetools points to customers holding full uptime through peak shopping events. It also highlights swapping out one component at a time, with no full rebuild needed.
Different vendors might have unique layers moving in the same direction. However, everything does converge on that purchase window.
How leaders can evaluate their own stack
For executives trying to assess where their own infrastructure stands, the guide gives a practical rubric that applies whether a business runs a consolidated platform or a composable assembly. Most organizations, it argues, already have the components. Few have the latency and integration required to use them together at the transaction moment. A few questions tend to expose the gap:
Can you actually activate customer data mid-transaction, or does it only kick in after the fact?
Session signals or last month’s profile, which one’s actually driving the decision?
Built inside the purchase flow as it happens, or locked in weeks ahead of time?
One placement, or the whole moment, selection, cart, payment, confirmation, all of it?
Value measured per transaction, or just at conversion?
Coordination- that’s the throughline running under every one of these questions we’ve addressed above. The companies actually pulling ahead aren’t the ones with the biggest toolkits. It’s the ones where data, decisioning, experience, and measurement all run as one low-latency loop, so every purchase does double duty: revenue now, and a signal that makes the next purchase sharper. That’s what the modern commerce stack is actually for, more than any single product or platform anyone’s selling.
