Ecommerce Personalization: The Unit Is the Visit, and a Shopper Is Not a Visit

Ecommerce personalization means changing what a shopper sees based on their data. The guides that rank for the term define it by the visit: six of the seven types Salesforce lists act on a page already open, and Bloomreach measures it as a conversion lift times traffic. The unit is the session, and a shopper is not a session.

Relvino decides per shopper rather than per visit, so read this as informed but interested. Three of the pages on the first page of results for this term are quoted directly: Salesforce’s guide, Bloomreach’s seven strategies, and Mastercard’s trends piece, published under its Dynamic Yield personalization brand. The argument about the unit is ours.

What ecommerce personalization is, in the guides’ words

Salesforce opens with a definition and later gives a second one. First: “Ecommerce personalization is tailoring the online shopping journey for each customer using data like past purchases and browsing behavior to display relevant content and product recommendations.” Then: “Ecommerce personalization is the practice of tailoring the online shopping experience to individual customers. This includes product recommendations, marketing messaging, and special offers personalized for each customer based on search and order history, and shopping trends over time.”

Bloomreach: “Ecommerce personalization is the practice of using data to understand your customers so you can offer experiences that are so relevant and contextual, they feel like magic. Ecommerce personalization is cross-channel and is driven by real-time, first-party data. It encompasses anonymous as well as known customers and includes messages, content, site layouts, products, and more.”

Both definitions are about the experience: what the shopper sees. Neither is about the prior question, whether this shopper should see anything from the brand today. That question sits in a different product category, and the gap between the two is the subject of this piece.

The seven types, and where each one runs

Salesforce lists “the major types of ecommerce personalization” as seven: product recommendations (“Those ‘You might also like’ suggestions”), personalized emails (“Tailored messages with items your customers have viewed, cart reminders, and offers based on their interests instead of generic blasts”), dynamic content (“Website elements that change based on who’s visiting”), personalized search results, customized landing pages, behavioral targeting (“Pop-ups and offers triggered by browsing behavior, like exit discounts or free shipping for hesitant shoppers”), and location-based personalization.

Count where each one happens. Recommendations, dynamic content, search results, landing pages, pop-ups and location-based content all change a page the shopper has already opened. One of the seven, personalized emails, reaches a shopper who is somewhere else. Six to one.

Bloomreach’s seven strategies lean further off-site: dynamic product recommendations, behavioral email marketing triggers, intelligent search personalization, location-based content, browse abandonment recovery campaigns, customer lifecycle messaging, and real-time content optimization. Four of the seven run on the site; three reach out. Even here the tilt is toward the visit, and the strategy it calls real-time content optimization carries a telling requirement: “This strategy needs a unified customer view across all touchpoints, typically powered by a CDP.” The page knows the shopper is one person across many visits. The product personalizes one visit at a time.

How the category measures itself

Bloomreach gives the formula outright, under “ROI Calculation Framework”: “Revenue Impact = (Personalized Conversion Rate – Baseline Conversion Rate) × Traffic Volume × Average Order Value.” Its first key performance metric is “Conversion rate lift: Track how personalized experiences improve conversion rates compared to generic experiences.” Algolia’s buyer’s guide for personalization platforms advises the same family: “Track metrics that tie directly to revenue: conversion rate lift, average order value (AOV), repeat purchase rate, and cart abandonment reduction.”

Every term in Bloomreach’s formula is a property of visits. Conversion rate is conversions per session. Traffic volume is sessions. Average order value is per order. A shopper who did not visit this month contributes nothing to any of the three, and therefore nothing to the revenue impact, however much that shopper would have bought with the right message on the right day. Of Algolia’s four, one, repeat purchase rate, has a shopper in the denominator. The rest are session metrics. This is not a flaw in the formulas. It is what they measure: the lift a personalization layer adds to the visits that happen.

The visitor the category cannot personalize

The vendors are candid about the hard case. Mastercard’s piece lists the obstacles to real-time personalization, starting with data: “Data issues: Limited data insights on new customers and an overwhelming amount of data insights for established customers.” Its first criterion for choosing a platform is the same problem from the buyer’s side: “Software that facilitates personalization for a first-time visitor to a site can partially compensate for the too-few-data-insights problem that new customers represent. The referring website is often a big clue as to the preferences of a first-time visitor.”

So the shopper with the least history gets the least personalization, by the category’s own account, and the best available clue is which site they came from. Bloomreach claims the anonymous visitor is in scope. Mastercard says what the scope amounts to: partial compensation, from a referrer.

Relvino’s starting point for that shopper is a prior learned across stores rather than a referrer: a Large Retail Model trained across 10K+ retailers and 7M+ data points carries what a first visit or a first order in a category tends to lead to, and the shopper’s own behavior takes over from there. It is one input among several, not the point of this piece; the point is that the category’s own experts place the first-time visitor outside what the visit alone can support.

Agentic, in the vendors’ words

Google’s related searches for this term now include “Agentic platform for personalization,” and Bloomreach’s site navigation calls its product exactly that. Its guide says: “Modern personalization increasingly relies on agentic AI that autonomously manages customer experiences in real-time, delivering relevance at scale rather than relying on hand-built rules.” And: “AI agents act as autonomous members of marketing and merchandising teams, handling data analysis, customer segmentation, and journey orchestration in real-time.” Salesforce: “AI makes it possible to deliver tailored experiences to thousands (or millions) of customers in real time.”

Read the nouns the agents are given. Data analysis, customer segmentation, journey orchestration. Segments are groups; journeys are paths a person designed; orchestration is sequencing inside them. An agent that does those three things faster is still working on the visit and the journey as units. The agentic marketing guide covers who owns the loop in that arrangement. For this piece the narrower point holds: the unit did not change when the adjective did.

From personalizing the visit to deciding for the shopper

A shopper is one person over months. They visit, leave, receive or do not receive a message, come back or do not, order or do not, and the gaps between those events carry as much information as the events. A system whose unit is the shopper reads all of it and asks a different question at every event: not what should this page show, but whether to act at all, on which channel (email, SMS/MMS or a pop-up on the site), with what offer, and when. Relvino makes that decision in under 80ms, per shopper, from live store events, inside guardrails the brand sets once: margin floors, quiet hours, channel limits, brand voice. The pop-up is where the two units meet: when the shopper is on the site, a pop-up is one of the actions available, decided for that shopper rather than rendered for that page.

The measurement changes with the unit. The registered results are stated per program and per customer, not per session: 2–6× ROI in 30 days as the headline claim, and per-customer figures such as Terra Kaffe’s 2X the revenue of standard flows. A brand that runs an on-site recommendation engine can keep it; what Relvino replaces is the layer that decides who hears from the brand, which on most stores is a set of flows nobody built around the shopper either. The lifecycle marketing guide covers that layer by name, and behavioral segmentation covers why the groups it runs on expire.

On-site personalization and per-shopper decisioning, side by side

In the list below, the personalization column is assembled from the guides quoted above; quoted phrases are theirs and the rest is a plain reading of the category. The decisioning column describes Relvino.

• Unit · On-site personalization, per the guides: The visit · Per-shopper decisioning: The shopper
• What it changes · On-site personalization, per the guides: What a visitor sees: recommendations, search results, layouts, pop-ups · Per-shopper decisioning: Whether a shopper hears from the brand, on which channel, with what, and when
• First-time shopper · On-site personalization, per the guides: “the too-few-data-insights problem that new customers represent”; the referrer as a clue · Per-shopper decisioning: A category prior from a Large Retail Model trained across 10K+ retailers
• Measured by · On-site personalization, per the guides: Conversion lift × traffic volume × average order value · Per-shopper decisioning: Revenue per shopper over time; 2–6× ROI in 30 days as the registered headline
• Rules · On-site personalization, per the guides: Hand-built rules, or “journey orchestration” by agents · Per-shopper decisioning: Guardrails set once; no journeys
• Example · On-site personalization, per the guides: Salesforce Commerce Cloud, Bloomreach, Mastercard Dynamic Yield, Algolia, by their own pages · Per-shopper decisioning: Relvino

For a brand already running a personalization platform, the test is to ask its dashboard a question it cannot answer: how many shoppers did not visit this month, and what did the platform do about them. For a brand deciding which layer to buy first, the Klaviyo vs Relvino comparison sets up a 14-day pilot beside the current setup, and pricing is public. The next best action marketing guide covers the enterprise decision layer that sits between a personalization engine and a person, AI agents for ecommerce sorts the agents by who they can reach, and the abandoned cart vs abandoned checkout guide is the case where the visit ends and the shopper continues.

Frequently asked questions

What is ecommerce personalization?

Changing what a shopper sees, on the site or in a message, based on data about that shopper. Salesforce defines it as tailoring the online shopping journey for each customer using data like past purchases and browsing behavior to display relevant content and product recommendations. Bloomreach calls it the practice of using data to understand your customers so you can offer experiences that are so relevant and contextual, they feel like magic, and adds that it covers anonymous as well as known customers. In practice most of it happens to a visit already in progress: recommendations, search results, page layouts and pop-ups.

What is the best e-commerce personalization software?

The pages that rank for that question are mostly written by the vendors themselves: Insider, Zoovu, Bloomreach, Algolia, Salesforce, BlueConic and Netcore each publish a list. Algolia’s buyer’s guide defines the category as software that leverages AI and machine learning to deliver tailored, personalized shopping experiences for each user across their customer journey, and suggests judging it on conversion rate lift, average order value, repeat purchase rate and cart abandonment reduction. The useful first question is which unit a tool optimizes: the visit, which is what on-site personalization engines are built around, or the shopper, which is what a lifecycle decision system is built around.

What are examples of ecommerce personalization?

Salesforce lists seven types: product recommendations, personalized emails, dynamic content, personalized search results, customized landing pages, behavioral targeting such as pop-ups triggered by browsing behavior, and location-based personalization. Bloomreach’s seven strategies overlap and add browse-abandonment recovery and customer lifecycle messaging. Six of Salesforce’s seven and four of Bloomreach’s seven change what a visitor sees while on the site; the rest reach a shopper who is not currently there, primarily through email.

What is AI personalization in ecommerce?

Personalization where a model, rather than a rule a merchandiser wrote, picks what each person sees. Bloomreach describes the current form as agentic AI that autonomously manages customer experiences in real-time, delivering relevance at scale rather than relying on hand-built rules. Salesforce says AI makes it possible to deliver tailored experiences to thousands or millions of customers in real time. The question to ask of any AI personalization product is what it decides: the content of a page a visitor already opened, or whether a given shopper should hear from the brand at all, on which channel and when.

How long does it take to migrate from a personalization and email stack to Relvino?

About 30 minutes for the technical cutover of the messaging layer: connect the Shopify store, point the sending domain, connect SMS, and order and browse history ingest automatically. An on-site recommendation or search engine can stay in place; Relvino decides per shopper whether to act, on which channel including a pop-up, with what offer and when, in under 80ms, and a new shopper starts from a prior learned by a Large Retail Model trained across 10K+ retailers. Revenue is then proven in a 14-day pilot run beside the current setup.