Two stores finished the month down twenty percent. One lost traffic. The other held its traffic and lost people somewhere between the product page and the payment form. On the revenue line those two months look identical, and the month of work that rescues the first store is a month wasted on the second.

So the question worth asking is not "is my conversion rate good?" It is "which part of my funnel got worse?" The first has no answer you can act on. The second does, and the numbers that answer it are already in your Shopify admin and cost nothing to pull. What costs something is reading them the way their labels invite you to, because the labels describe one thing and the arithmetic underneath counts another.

The five numbers, in three tiers

Open Conversion rate breakdown in your admin, where Shopify's documentation gives its four steps these labels: All sessions, Sessions with cart additions, Sessions that reached checkout, Sessions that completed checkout. Sort what you find into three tiers.

  • Traffic — total sessions.
  • Funnel — the share of sessions with a cart addition, the share that reached checkout, the share that completed checkout. That last one is your conversion rate, not a sixth number.
  • Basket — average order value, from Shopify's own tile.

Every funnel share is out of all sessions — the bar lengths are a shape, not data. Labels follow Shopify's Conversion rate breakdown report, read 13–14 August 2026.

The tiers are the three things revenue is made of: sessions × conversion rate × average order value — the same multiplication IRP Commerce builds its own turnover estimates from. Ten per cent on any one of them is worth ten per cent on either of the others. Don't expect the three to multiply back to last month's takings to the pound, though — a customer can buy twice in one session, and the tile leaves shipping and tax out. Use the multiplication to point you at a stage, not to check your books with it. The funnel tier is that middle term split into stages you can read.

One line of Shopify's documentation decides whether you read them right. "The conversion rate for each step is calculated as the number of sessions for a step divided by the number of total sessions." Every percentage on that report is out of everybody who visited. "Reached checkout" does not mean "of the people who added to cart", so you can compare any two steps straight off, with no maths in between. One setting still decides what those middle numbers mean: the report carries an Open Funnel toggle. Open, it counts a session that jumped straight to checkout without ever touching the cart; closed, it counts only the sessions that walked every step in order. Shopify's page does not say which your report starts in.

The basket tier has a definition problem sitting on your own screen. Shopify's tile is built on gross sales, so it leaves shipping and tax out and takes the discounts off. Divide Total sales by orders yourself and you will usually get a bigger number, because that figure adds shipping, tax, duties and fees and takes off refunds. Use the tile.

Then pull one full calendar month of all five. A month, not a week, for a reason two sections down.

How to read the shape

Five numbers give you a shape rather than a score, and the ruler is your own history. Last month tells you what has just moved. The same month a year ago tells you what moved once your season is allowed for. Then read the tiers against each other.

Traffic holding, cart additions down on last year. The loss is before the decision, on the product page and the offer sitting on it.

Cart additions holding, fewer sessions completing checkout. The leak is after the decision — inside checkout, or inside what checkout reveals about cost and delivery. Read that label before you trust the gap. "Sessions that reached checkout" counts sessions where there was "user input (for example, a key press or mouse click) during checkout" — not sessions that arrived at checkout. Somebody who lands there, reads the shipping cost and closes the tab without touching anything is not in that number. She sits in the gap between cart and checkout, indistinguishable from people who never went that far.

Everything steady, the average order falling. Not a leak at all; it is what one order contains.

None of that tells you what to change. It tells you which page to open, which is the one thing the revenue line will never do.

Why a weekly number lies at your volume

If you take 40 orders a month, one order either way can make an ordinary week look like a crisis. Work it through: 40 orders on 2,000 sessions is 2.0%, about ten orders in a normal week on about 500 sessions, so one order more or fewer moves a weekly conversion rate by roughly 0.2 points. A week reading 1.6% against a week reading 2.2% is three orders of ordinary lumpiness. Read months.

The published numbers, as context

IRP is the only one that answers all three questions you should ask any benchmark: how it did the sum, whose shops, and which month — though never how many merchants. Littledata's are the only numbers here counted on Shopify stores specifically, which is why they earn a row despite being from 2023 — and the device split matters if your traffic is phones: 1.2% mobile against 1.9% desktop.

Baymard's 70.22% is an average of averages, and the page carries no definition of abandonment, because each of the fifty studies brought its own. Baymard sells checkout auditing, and many of the fifty come from firms selling cart-recovery software. Everyone publishing a number here has something to sell, ourselves included: we sell page-building software, this community is how people find it, and we have no funnel measurement of our own.

Your own vertical is the one place a number does exist. Salesforce's Shopping Index publishes an online conversion rate for sixteen named retail verticals, three of them beauty — running from 2.4% for Beauty, Skincare down to 1.3% for Beauty, Makeup, against 1.4% across all verticals. Ask it the three questions and it answers half of one. If you sell skincare, 2.4% against the 1.4% all-vertical line is worth a glance — held loosely. Who was measured is stated only in an April 2024 Salesforce blog post — the index runs on Salesforce platform data, though Salesforce adds there that the figures are extrapolated to the wider retail industry and "not indicative of Salesforce performance". What it divided by is not on the dashboard, nor on the two Salesforce blog posts that carry the method. No source here names a vertical for supplements or fitness, and IRP's nearest is a broad Health and Wellbeing retail category.

Average order value is where a benchmark stops meaning anything. Inside a single source, one month and one formula, IRP's ten markets for June 2026 run 14.5-fold apart. A £55 wellness order and an £800 nursery order cannot share a benchmark.

The same word also covers different sums: IRP and your Shopify dashboard both divide by sessions, Dynamic Yield divides by users, and Contentsquare counts visits and treats a form submission as a conversion. On 13 August 2026, Dynamic Yield's own benchmark page was telling the world that "The average eCommerce conversion rate globally is 0%". The page is broken, and Shopify's conversion-rate article, updated 4 February 2026, still sends readers there for its 2.95%.

So: your own vertical's number held loosely, Littledata's device split if your traffic is phones, and the rest as context.

The comparison that is actually yours

Every figure above is somebody else's shops counted somebody else's way, and none of it can tell you what will happen if you change something on yours.

So write these down for last month, and put the same month a year ago beside them:

  • Total sessions
  • Sessions with cart additions (share)
  • Sessions that reached checkout (share)
  • Sessions that completed checkout — your conversion rate
  • Average order value from the tile

Let the difference choose the page you open this week. For a second pair of eyes, post them in the community, and say whether the report was set to open or closed funnel: the toggle sits on the report itself. Post the shares and say whether traffic and basket rose or fell; keep the raw counts to yourself if you'd rather. We will read them back with you. Nobody needs your store's name to read a funnel.