Conversion Rate Denominators and Why They Differ by Platform

Every conversion rate published anywhere runs on the same arithmetic: orders divided by sessions, multiplied by 100. Nobody disputes the formula. What varies, platform to platform and report to report, is what counts as a session in the first place, and that disagreement changes the output even when the underlying business hasn't changed. A single shopper who visits a store three times in one week counts as three sessions under one counting method, or as one visitor under another, and a platform that reports by unique visitors will post a higher conversion rate than one reporting by sessions, with no difference in how many orders actually came in. Vantainsights' 2026 guide puts the point directly: firms that report by unique visitors will show higher conversion rates than those reporting by session, even with identical revenue. If two conversion rates don't match, that isn't a sign that someone's tracking is broken. It reflects two different, equally legitimate definitions of the denominator, built into how each platform's architecture decides what to count as a visit in the first place.

How Shopify counts sessions differently from GA4

Shopify Analytics and GA4 define sessions differently even when pointed at the same storefront, which makes the numbers they surface structurally incompatible. Both tools look at identical traffic and identical orders, but they're built to measure different things, so the conversion rates they produce are structurally incompatible rather than simply inconsistent. Shopify calculates its online-store conversion rate as sessions that completed checkout divided by total sessions, and because a single session can contain more than one purchase, orders and purchasing sessions can drift apart even inside Shopify's own counting. GA4 works from an entirely different session model: a session resets when the traffic source or medium changes, at midnight, and after a stretch of inactivity, none of which maps cleanly onto how Shopify defines a session boundary. A brand running both tools side by side will routinely see two different conversion rates, and the right response is to recognize that the tools are answering two different questions about the same traffic, not to figure out which one is wrong.

Denominator definitions across Adobe Commerce, Salesforce Commerce Cloud, and WooCommerce

Platform-level conversion benchmarks for Adobe Commerce, Salesforce Commerce Cloud, and WooCommerce are mostly agency-derived or inferred, which makes cross-platform comparison doubly unreliable: the denominators differ and the sample quality differs. Elogic's 2026 benchmark report finds that platform-level conversion data is much weaker than people assume, and while Shopify has the strongest published benchmark set, the published figures for Adobe Commerce, Salesforce Commerce Cloud, and WooCommerce are mostly agency-derived or inferred rather than drawn from a comparable first-party dataset. WooCommerce illustrates the gap concretely: its own Analytics → Revenue report tracks gross sales, returns, coupons, net sales, taxes, shipping, and total sales, with no conversion rate metric in the report at all, and because WooCommerce runs on WordPress, the underlying session definition depends on how a given install handles cookies and plugins, which varies by implementation in ways a packaged platform like Shopify doesn't have to contend with. The same 2026 guide groups Shopify into one average conversion range, but it places WooCommerce and Magento or Adobe Commerce into different ranges entirely, and it cautions that implementation quality affects conversion outcomes more than which platform a merchant happens to run. When an agency reports a client's Adobe Commerce conversion rate and sets it beside a published Shopify benchmark, the comparison can be apples-to-oranges on two separate axes at once, the denominator definition and the sample behind it, rather than just one.

Why email conversion rate benchmarks are frequently misread

Denominator differences extend to the platform layer as well. One is session-based conversion rate: clicks from an email that lead to a purchase, counted as a share of the sessions those clicks produced. The other is placed-order rate: purchases counted as a share of everyone who received the email, whether they opened it or not. Both numbers are real and both get published as "email conversion rate." Elogic's 2026 report names the confusion directly: email is the highest-converting channel, but its benchmarks get misread constantly because sources mix session-based conversion with placed-order rate per recipient. The ranking only holds if the denominator being used is sessions generated by email clicks, not the full list of recipients the email went out to. If you check your email conversion rate against a published benchmark without first confirming which denominator it used, you're set up to draw the wrong conclusion about how well the channel is performing.

Amazon's conversion rate versus a DTC storefront's

Marketplace conversion rates belong in a different category, and Amazon shows this most clearly. Amazon's conversion rate runs high not because its shopping experience outperforms a DTC storefront, but because its denominator excludes almost the entire top of the funnel that a direct-to-consumer brand has to pay to build and then measure. A shopper who lands on Amazon has usually already typed an intent-laden search query into a buying environment, so the research and consideration happened somewhere else, before the session Amazon counts ever began. A Shopify session, by contrast, can start on a blog post, a paid social ad, or an informational landing page that has nothing to do with imminent purchase intent, so the denominator for a DTC brand includes a mass of traffic that was never close to buying anything. If you set a DTC brand's session-based conversion rate beside Amazon's, you're comparing a funnel that starts at awareness to one that starts at purchase intent, and no amount of on-site optimization changes which funnel a brand is running. That distinction closes out the platform-by-platform survey and points toward a bigger question: once sessions, visitors, and placed-order rates all mean different things depending on where they're measured, what does the published spread of conversion benchmarks actually tell a brand trying to compare itself to the field?

Denominator variance across the published 2026 benchmark spread

Published 2026 conversion benchmarks span a wide range, running from figures well below the commonly cited midpoint to numbers nearly a full percentage point above it, for businesses that look, on paper, comparable. That spread is mostly a function of denominator differences rather than one group of merchants genuinely outperforming another. The sources behind these benchmarks don't disagree about basic facts; they're built from different denominators, sessions in one case and visitors in another, from different samples such as IRP's UK-weighted merchant panel versus Shogun's pool of active Shopify stores, and from different time windows, a single month in one report, rolling 12 months in another. Speed Commerce's 2026 analysis adds a geographic wrinkle: IRP's dataset covers B2C merchants trading in Great Britain, Northern Ireland, and Ireland, and a UK shopping basket isn't a US one, so currency conversion moves the reported dollar figures independently of how the underlying businesses actually traded. Industry segmentation compounds the spread further. Elogic's 2026 report shows a wide global band once industry is factored in, with Food & Beverage at the top and Luxury & Jewelry at the bottom. Speed Commerce's own June 2026 category data shows Arts and Crafts converting at the high end and Baby and Child at a fraction of that figure, so any single cross-category average is close to meaningless as a planning number. Put together, these figures mean a benchmark cited without its denominator definition, its sample composition, its geography, and its time window attached isn't a usable benchmark: it's a number waiting for someone to supply the context that would make it mean something.

A new denominator problem no platform currently handles: AI-referred sessions

Every disagreement covered so far involves counting methods that already exist inside known systems. The next one doesn't fit inside any of them. When a shopper uses ChatGPT or Perplexity to research a purchase, the comparison shopping, the reading of reviews, the narrowing down of options, all of it happens inside that conversation, so by the time the shopper clicks through to a store, they aren't browsing anymore. They've largely already decided. That changes what a "session" from an AI referral actually represents compared with a session that starts on a cold paid-social ad. In dollar terms the shift is still small: across the same set of stores, Shopify's own Shop app continues to out-earn every AI chatbot combined by a wide margin, even though the volume behind AI-referred traffic remains modest, the conversion quality it brings is genuinely higher. Attribution compounds the undercount. A shopper who does the deciding with ChatGPT's help but completes the purchase later, during a sale, gets logged as a direct or email conversion in most analytics setups, so whatever share of sessions gets labeled "AI-referred" today is a floor on how much influence AI research actually had, not a ceiling on it. If AI-referred sessions convert at a meaningfully higher rate and a brand has no way to isolate them inside its own reporting, blended conversion rates will drift upward modestly over time, and the gain will get credited to whatever channel happens to sit next to it in the attribution chain, rather than to where the decision actually got made.

The next stage of this problem is structural rather than attributional. If AI buyer agents start making product selections for a shopper instead of just helping them research, the session-based conversion denominator can't describe what happened, because the whole consideration process takes place where a brand's analytics can't see it. OpenAI deprecated Instant Checkout in March 2026, and the model that replaced it, ACP, handles product discovery and then redirects the shopper to the merchant. Brands keep the login data, the loyalty engagement, and the direct customer relationship on their own site rather than losing it to an in-chat transaction. Roughly 30 Shopify merchants, including Etsy, Glossier, SKIMS, Spanx, and Vuori, had gone live on ChatGPT's Instant Checkout before it was pulled in March 2026, and their products are now surfaced for discovery and redirect rather than completed purchase inside the chat window, with Walmart and Target currently live as partners under the newer model. Amazon's Alexa for Shopping already serves a large user base, and the estimated incremental sales figure cited for 2025 is a measurable revenue line, substantial enough to make clear that agentic-assisted commerce isn't a future scenario. The metric that would describe this channel is a conversation-to-conversion rate, the share of engaged agent conversations that end in a purchase, but no analytics platform currently surfaces that number natively for a brand's own store. Understanding the denominator becomes especially important for DTC brands adding a new revenue channel on top of an existing storefront. When a brand layers an AI sales surface onto its site, the session definition feeding conversion tracking has to stay the same between the baseline period and the post-deployment period, or you can't trust any reported lift. Without that consistency, a brand's reported conversion rate reflects only the last click a shopper made, not the decision that preceded it, and optimizing against that number means optimizing against a measurement that no longer describes where the sale was actually won.