Real-time relevance is the practice of reading live signals from a single purchase (what the shopper just selected, the value of the order, the device in hand, and prior behavior) and using them to pick the best next action for that one person in that one moment. It is the distinction at the center of how Rokt, the New York-based e-commerce technology company, describes the future of AI in commerce in a recent analysis. Real-time relevance, the company argues, is a different thing from personalization, and it is starting to outperform it. Personalization sorts shoppers into segments and serves each group pre-assigned content. Real-time relevance decides for the individual, live, at the transaction.
Rokt’s larger claim is about where that decision matters most. The window around the purchase itself, from product selection through cart, payment, and order confirmation, is the highest-value surface in e-commerce, because attention, intent, and trust all peak there. An offer that fits that state reads as a service rather than an interruption. Rokt reports a 4.03 percent click-through rate across what it calls Transaction Moment placements, roughly ten times the rate it cites for Google Display, a gap it attributes to meeting intent already in motion instead of interrupting it.
What real-time relevance means, and why it is not personalization
Personalization has been the industry’s default for a decade. It groups customers into cohorts and assigns each cohort content, an improvement on mass messaging but limited by design, since individual behavior diverges from the group average constantly. Real-time relevance, in Rokt’s framing, works from live transaction signals rather than fixed segments, and it targets the individual instead of the cohort. The argument rests on a technical point that recommendation researchers broadly accept: models trained on individual-level interaction data capture variance that cohort averages erase, so they predict the next best action more accurately. A platform reading tens of millions of transactions a month has far more signal per person than one leaning on segments, and that edge compounds as volume grows.
Why the transaction window is e-commerce’s most underused surface
Rokt breaks the window into three moments, each with its own logic. The cart and checkout stage is high-intent but time-pressured, so offers have to render almost instantly and take one tap to accept, which is why upsells and complementary products fit there. The payment stage runs on trust, since the shopper is entering sensitive details, so payment-method incentives and loyalty activations sit naturally alongside the financial decision. The post-purchase confirmation stage is calm and uncrowded, with no abandonment risk left, which makes it the place where subscriptions, ancillary services, and third-party recommendations earn attention no pre-purchase surface can match. Static, generic content in any of those moments, Rokt argues, is revenue left on the table.
How AI changes the economics of the transaction
Moving from rule-based offer logic to AI decisioning changes the math in three ways, by Rokt’s account. It removes what the company calls the relevance tax, the weaker conversion and added friction that come from showing everyone the same offer regardless of fit. It rewards precision over volume, since a system that selects one genuinely relevant offer tends to beat one that surfaces a dozen adequate ones, which puts the value in selection accuracy rather than option count. And it compounds, because every offer shown and every outcome recorded feeds a closed-loop model that sharpens the next decision. Platforms built on transaction-level AI tend to widen their accuracy edge over time instead of plateauing.
First-party data is both the constraint and the advantage
Every relevance system runs into the same tension: the data that makes relevance possible is the data shoppers guard most closely. Regulations including GDPR and CCPA limit how behavioral data can be gathered and used, and the fading of third-party cookies has pushed the industry toward first-party data. Rokt’s position is that this is not a tradeoff. A customer’s own purchase history, current cart, and live session behavior are richer predictors than any third-party profile, and using them requires no cross-site sharing, which makes first-party data both the most compliant and the most powerful input available.
Rokt put money behind that view. In January 2025, it made a roughly $300 million investment in mParticle, a customer data platform, and folded a real-time first-party data engine into its relevance stack. Analysts read the move as a bet beyond advertising. The research firm Sacra noted that Rokt was expanding across platforms and accumulating first-party data to drive timely offers from first visit through checkout and post-purchase, and CDP.com placed the deal within a wider wave of customer-data-platform consolidation in early 2025. Co-founder and chief executive Bruce Buchanan framed the merger as a way to bring a performance lift to every client, while mParticle chief executive Michael Katz described the combined offer as letting brands “activate their data in real time” while keeping full ownership of it. Rokt says joint clients had already seen up to 50 percent better consumer and business outcomes before the deal, and mParticle’s more recent releases include a matching tool the company says lifts identity match rates by more than 25 percent.
The next phase is agentic commerce
Real-time relevance at the transaction is the current state of the art. The next phase, in Rokt’s telling, is agentic commerce, where AI systems start or complete purchases on a shopper’s behalf through a chat interface, a voice assistant, or a standing rule like a price drop or a replenishment window. The forecasts differ by definition but not by direction. A December 2025 eMarketer forecast puts AI platforms at about $20.6 billion of US retail e-commerce sales in 2026, rising to $144 billion by 2029, and Morgan Stanley’s broader estimate reaches $190 billion to $385 billion in US e-commerce sales by 2030. Rokt’s point is that the same transaction-window relevance that wins today is the infrastructure those agentic systems will run on tomorrow, so the work is not optional even for merchants who see agents as years away.
Why Rokt’s read carries weight
Rokt has spent more than a decade at the point of purchase, which gives its view some standing. Its AI engine, the Rokt Brain, weighs each transaction in real time against more than 1.95 trillion data points a year, and, per Rokt by the Numbers, the network will power more than 10 billion transactions in 2026, reach 165 million monthly active users, and serve over 33,000 clients, including more than half of the leading e-commerce companies globally. In April 2026, the company was named for the first time in Gartner’s 2026 market guide for retail and commerce media networks, where chief revenue officer Craig Galvin called the transaction “one of the most valuable and underleveraged” interactions in commerce. Speaking at an eMarketer summit, Rokt’s Sophie Donoghue made the same case from the shopper’s side, noting that the clearest sign someone is open to an offer is “when they’re literally buying something.”
The outside record tracks the argument. Rokt’s revenue passed $800 million in 2025, and the company landed on the 2025 Deloitte Technology Fast 500, milestones noted by outlets including Dataconomy in its analysis of the company’s growth. The Silicon Review described Rokt as the firm that has done the most to articulate the idea that the transaction, not the ad or the search result, is the highest-value moment in commerce, and its Gartner debut arrived as commerce media accelerated, with US retail media spend projected to reach $69.33 billion in 2026, up from $58.79 billion in 2025 on eMarketer figures.
What it means for merchants
Rokt’s guidance for merchants comes down to four moves:
- Treat the transaction window as a product surface, not a handoff, and give the cart, payment, and confirmation pages the same design discipline as any other part of the experience.
- Invest in closed-loop attribution at the individual level, since relevance models only improve when fed precise outcome signals.
- Build for agentic readiness now, because merchants who have already optimized the transaction window for relevance will translate more easily into agent-driven formats.
- Design data strategy around first-party signals, the highest-quality input for real-time relevance and the most durable asset in a privacy-constrained market.
The merchants who pull ahead over the next five years, Rokt argues, will not be the ones with the biggest catalogs or the lowest prices. They will be the ones that treat every transaction as a chance to be useful and build the infrastructure to act on it in the moment. For its part, Rokt continues to publish its view of the shift, and the throughline holds: as AI moves deeper into how people buy, the advantage goes to whoever can read intent in real time and answer it with something the shopper actually wants.
