When your conversion rate moves and your shoppers do not

Open your Shopify analytics this week and the conversion rate may not match last week's. Before you change anything on the store, check whether your shoppers changed or the measurement did.

Shopify rolled out a session measurement update between 21 and 23 September 2026. Sessions now end after thirty minutes without activity instead of at midnight UTC. Sessions that carry no pageview, such as a checkout opened straight from a saved cart, are counted for the first time. Bot traffic is filtered out of session reports by default.

Orders, sales and customer counts are not affected. What moves is everything built on top of sessions: conversion rate, add to cart rate, reached checkout rate, bounce rate.

A smaller denominator is not a better store

Conversion rate is orders divided by sessions. Change the way sessions are counted and the ratio moves without a single shopper behaving differently.

Filtering bots removes visits that were never going to buy, which pushes the rate up. Counting sessions that carry no pageview adds visits the report used to miss. Either direction is a definition changing, not a decision a customer made.

Shopify's own guidance is to read the session change against orders and sales. If the orders held steady while the percentage moved, your store did not change.

Rebase before you compare

Historical data is not reprocessed. A chart that runs across the rollout puts two different definitions on the same line and invites you to read the join as a trend.

Treat the days after the update as a new baseline, and write that date on any report you circulate. A September to October comparison is then read as what it is.

The figures that do not depend on sessions stay comparable across the join: orders, revenue, and average order value. If you need a like-for-like read of the quarter, start there.

The rate never told you why

Even a perfectly counted conversion rate says how many sessions ended in an order. It does not say what stopped the rest.

In the report, a shopper who left because a measurement was missing from the page looks exactly like a shopper who was only browsing. One of those is a problem you can fix this week.

The reasons come from somewhere else: the questions your support team hears, the notes customers leave on returns, the sentence someone types before closing the tab. The cost of a question nobody answers covers how to collect them.

Watch what a definition cannot move

Pick a few measures that sit outside this change and follow them through the quarter. Orders by product line. Revenue net of returns. Repeat purchases from the same customer. The share of orders where someone answered a question first.

They are slower to read than a percentage on a dashboard, and much harder to move by accident.

Live shopping belongs on that list. When an advisor takes a call and the customer buys, the result shows up in the order and in what the customer chose, not only in a session count that a platform can redefine next year.