Definition
Ecommerce conversion rate is the share of visits that end in a purchase, calculated as orders divided by sessions for a period and expressed as a percentage. Every reporting tool defines both halves slightly differently, which is why a store can see three conversion rates for the same day in its admin, in GA4 and in an ad platform and find that all three are correctly calculated.
Formula
Conversion rate = (Orders ÷ Sessions) × 100
- Orders
- Completed purchases in the period from one source, either the store's order records or the analytics purchase event, not both
- Sessions
- Visits counted by the same source, using its own rules for when a visit starts, times out or restarts on a new campaign
One week seen from the store admin and from GA4
- Sessions in the store admin
- 18,400
- Orders in the store admin
- 331
- Conversion rate in the admin
- 331 ÷ 18,400 = 1.80 percent
- Sessions in GA4 for the same week
- 19,900
- Purchase events in GA4
- 318
- Session conversion rate in GA4
- 318 ÷ 19,900 = 1.60 percent
- What explains the gap
- Different session rules, consent choices, blocked scripts and orders placed after the analytics session expired
Illustrative figures. Neither tool is broken. Pick one as the reporting source of truth, keep the other for diagnosis, and never move a target from one to the other mid-quarter.
Why it matters
Conversion rate matters because it is the only lever in the revenue equation that costs nothing in media. A store that keeps its traffic and improves the path from product page to receipt earns more from the same spend, and the gain applies to every channel at once. It is also the fastest diagnostic the store has. A drop that appears on one device, one browser, one country or one template is usually a specific broken thing, such as a payment method failing validation or an app slowing the cart, and the segment where it appears names the cause. Read alone it is a vanity figure, but read by segment, next to order value, it points at work worth doing.
Where it goes wrong
- Chasing a published industry average: catalogue, price band, traffic mix and whether the store runs brand campaigns move the figure more than any site change, so the honest benchmark is the store's own trend
- Reporting one site-wide number when the device mix has shifted: more phone traffic lowers the blended rate even if the phone experience improved, which is why device is separated before conclusions are drawn
- Counting sessions that could never buy: staff visits, uptime monitors, scrapers and traffic from countries the store does not ship to all sit in the denominator until they are filtered
- Comparing a promotion week with a normal one: discount traffic converts on a different curve, and a rate that fell after a sale usually reflects who arrived rather than how the site performs
- Treating a rise as proof a test worked: cutting spend on prospecting raises conversion rate while shrinking orders, so the count of orders is read next to the rate every time
Questions about ecommerce conversion rate
Should conversion rate be measured per session or per user?
Per session for site and merchandising work, per user when judging a channel that people research over days. A shopper who visits four times before buying produces one conversion in four sessions and one conversion in one user, so the user-based figure is always higher. Both are legitimate; the reports simply have to say which one is on the chart.
Why is the conversion rate in Google Ads higher than in my store admin?
Google Ads divides conversions by clicks and counts a conversion inside its own attribution window, including orders placed days after the click and, depending on settings, orders following a view. The store admin counts orders against sessions on the site in the period. Different numerator, different denominator, and neither is a check on the other.
How much traffic does it take to trust a change in conversion rate?
Enough orders, not enough sessions. A store with a low order count sees the rate swing on normal weekly variation, so a fortnight of comparable traffic and a few hundred orders is a reasonable floor before calling a movement real. For deliberate experiments, the sample is calculated in advance from the current rate and the smallest difference worth acting on.