Ecommerce Conversion Rate: Formula, Benchmarks and Fixes
read
·

Ecommerce conversion rate is the share of sessions on your site that end in a purchase. Divide orders by sessions and multiply by 100, so 360 orders from 12,000 sessions gives a 3.0 percent rate.
That's the arithmetic. Everything else in the conversation is interpretation, and most of the interpretation you'll read is wrong. The popular advice says to push your rate toward an industry average of two to three percent and treat the gap as a problem with your website. It's tidy advice, easy to repeat in a meeting, and it hides the fact that the same 3.0 percent can be excellent for one traffic cohort and poor for another.
A rate isn't a score. It's a ratio with a denominator you chose, a numerator you filtered, and a device and channel mix you inherited from your ad account. Once you see it that way, a falling rate stops being a mystery and becomes a diagnostic question with a specific answer: which stage broke, and what's the first thing to check there.
Table of Contents
How to Calculate Ecommerce Conversion Rate
Think of a shop window on a busy street. If 100 people walk past and 3 come in and buy, the rate is 3 percent. Your analytics platform counts the walk-past as sessions and the purchase as an order.
The calculation takes three steps, done in one tool.
Pull both numbers from the same source. Same date range, same filters, same timezone, same exclusion of internal traffic.
Divide orders by sessions. 360 orders divided by 12,000 sessions is 0.03.
Multiply by 100 and round to one decimal. That gives a 3.0 percent ecommerce conversion rate.
The full formula in one line is (360 / 12,000) x 100 = 3.0. Now change only the traffic. The same 360 orders from 9,000 sessions is 4.0 percent. From 18,000 sessions it's 2.0 percent. Revenue never moved. The denominator did.
The relationship between the three numbers is worth keeping in front of you while you read any benchmark.

Where the Calculation Usually Breaks
Most wrong rates come from mixing platforms. Sessions pulled from one analytics tool, divided by orders pulled from the store admin, with two different attribution windows, two cookie policies and two timezones. Neither number is wrong. The pair is.
Strip test orders, cancelled orders and refunded orders you've decided to treat separately before you divide, and do it consistently. If you remove cancellations in March, remove them in February too.
Sessions and Users Are Different Denominators
A shopper who visits three times over three days counts as three sessions and one user. Divide your orders by users and the rate rises. Divide the same orders by sessions and it falls. Both results are honest as long as you label which one you used.
For ad-driven stores, sessions are the better denominator because they match how paid traffic arrives. A single click is a session, and a session that doesn't buy is the thing you're trying to fix.
Check before trusting a trend: recalculate last month by hand, on a calculator, and confirm the result matches your dashboard. If the two disagree, fix the dashboard before you read any month-over-month movement.
Ecommerce Conversion Rate Benchmarks by Source and Device
A 2% rate can be a strong month or a weak one. Device, traffic source, and denominator decide which. A published figure only helps when its scope matches the rate you calculated.
Source | Figure | Scope |
|---|---|---|
Contentsquare, 2026 | Desktop 3.4%, mobile 2.0% | Site-level sessions by device |
Dynamic Yield | Mobile 2.86%, desktop 2.23% | Site-level sessions by device |
Littledata Shopify sample | Mobile median 1.6%, desktop median 1.5% | Shopify store device split |
Littledata, 450 stores | Median 1.3% | Shopify stores, 90 days ending November 2026 |
CartFlows, rolling 12 months | Global average 2.74% | Completed purchases per visit |
ConversionTeam, 2026 | Median 3.9% | Site-level, 456 A/B tests across 31 clients |
Unbounce, year ending July 2024 | Median 6.6% | Landing pages across industries |
Contentsquare reports 3.4% on desktop against 2.0% on mobile, while Dynamic Yield reports 2.86% on mobile against 2.23% on desktop, as collected in this benchmark roundup from ClickPost. The device direction flips. One dataset says desktop converts better. The other says mobile does.
The Littledata Shopify sample's medians, 1.6% on mobile and 1.5% on desktop, are lower than both the Contentsquare and Dynamic Yield device rates above. A separate Littledata analysis of 450 Shopify stores over the 90 days to November 2026 put the median at 1.3%.
Wider samples sit on still different scopes. ConversionTeam's 2026 analysis found a median site-level rate of 3.9% across 456 audited A/B tests involving 31 ecommerce clients. CartFlows reports a global average of 2.74% in 2026, measured as completed purchases per visit over a rolling 12-month window. Unbounce's benchmark data puts the median landing-page rate at 6.6% across industries for the year ending July 2024.
Why the Same Benchmark Points in Opposite Directions
Those sources differ on four things at once: the denominator, the traffic mix, cookie consent rates, and how each one handles bot traffic and app sessions. A store with heavy returning traffic and a strong email list will beat a store of the same quality that buys cold social clicks.
The landing-page figure shows the scope problem most clearly. A landing page receives people who already clicked an ad for one specific product. A sitewide rate also counts every visitor who left from a blog post. Comparing the two is like comparing a checkout queue to a shopping mall. The first group already chose a product. The second group is still wandering.
Pick the Benchmark Whose Traffic Matches Yours
Match the cohort before you match the number. Compare your mobile rate to mobile figures, your paid social rate to paid social figures, and your returning-visitor rate to returning-visitor figures. Then stop treating the industry average as a target. Use your own trend, split by the same device and source, as the standard.
If the blended rate falls, check the mix before you rewrite the offer. Stores that lean on cold social traffic sit in a different cohort, and how short-form video traffic converts shows why that mix pulls the sitewide rate down without meaning the product page or the checkout is broken.
A brand running Meta or TikTok ads can show a lower sitewide rate than a search-led competitor and still have the healthier funnel. Cold, mobile-heavy visitors convert later, often on a second or third visit.
Checkout Friction That Costs Conversions
A shopper who already decided to buy can still leave at the form. 17% of US online shoppers abandoned an order because the checkout was too long or complicated, the average checkout displays 23.48 form elements by default, and usability testing produced a workable flow at roughly 12 elements (seven fields, two checkboxes, two dropdowns and one radio-button interface), according to Baymard's checkout usability research.
Roughly half the default form is optional.
That is the first check at this funnel stage, not a hunt for an industry average. A 2.5% sitewide rate and a weak checkout completion rate can describe opposite problems. One counts everyone who landed, including people who never meant to buy. The other counts people who wanted the product and then stalled on the form. Device and traffic source split them again: the same field list punishes a cold mobile click harder than a returning desktop visit.

The mechanism is cumulative interaction cost. Every extra field adds typing, validation, a moment of hesitation and one more chance to fail on a phone keyboard. Across a default form of about 23 elements, that cost is why a shopper who wanted the product twenty seconds ago closes the tab.
Mobile shoppers and high-intent visitors arriving from paid social take the hardest hit. A small screen and a short patience window turn a field desktop users skip into the step where the order dies. Split the drop by device and by source before you rewrite the offer.
Fix the Form Before You Fix the Messaging
Work in this order, because the first two steps decide whether anything downstream matters.
Instrument by step and error type. Log where the drop happens and which error fires, not just the total abandonment count.
Cut fields you don't need to ship or charge. Billing address for a digital add-on, a second phone number, a company name nobody uses.
Show shipping and tax before payment. Surprise costs at the final step are their own abandonment category.
Test guest checkout. If account creation is mandatory, you're charging a registration fee in attention.
If you're rebuilding the page itself, start by browsing checkout tools that boost AOV before redesigning, since the field list and the payment options come out of the same decision.
Cart Recovery Asks What Stopped the Order
Exerta offers abandoned cart recovery by AI phone call plus SMS, with iMessage for the highest-value carts on iPhone. SMS serves other shoppers, while calls can address big carts and failed payments. The point is to ask what stopped the order rather than assume it was price.
A failed card, a form validation loop and an unanswered sizing question look identical in an abandonment report. They need three different messages. Send the original cart context back, not a generic nudge, and don't auto-apply a discount before you know the cause.
For US marketing texts and AI or prerecorded voice calls, you need prior express written consent under the TCPA, a clear disclosure beside the phone field with no pre-ticked box, STOP honoured and HELP supported, sends only between 8am and 9pm in the recipient's local time with stricter state rules observed, and applicable A2P/10DLC registration for texting. A phone number on its own isn't consent. This isn't legal advice, so confirm your setup with counsel before you turn campaigns on. Read the wider ecommerce customer experience as one system rather than a checkout problem and a support problem.
Reading Social and AI Referral Traffic Separately
A sitewide conversion rate treats a TikTok click and an AI-referred visit as the same shopper. They are not. Adobe's analysis of more than one trillion visits to U.S. retail sites found generative-AI referrals were 23% less likely to convert than other traffic in July 2025, after gaps of 49% in January and 38% in April, as Digital Commerce 360 reported.
Same source, six months, a shrinking gap. Shopper intent moved through a research cycle. An industry average would have called the channel weak in January and only soft by July. Both readings describe the blend. Neither tells you which page, device or denominator moved.

Build Cohorts Instead of One Number
Split AI-referred and paid social visitors by landing page, product category, device, assisted conversion and time to purchase. Compare each cohort with its own previous period, not with the sitewide rate.
A shopper can read an AI comparison in week one, search the brand in week two and buy from an email in week three. A last-session model books that order as email and marks the AI visit a miss. Getting conversion attribution right is what shows the assisted revenue those earlier sessions produced.
On mobile, the same cohort often looks worse because the landing page is slower and the product choice is still open. On desktop, a returning visitor from a brand search can post a high rate on a short path. Averaging those two paths produces a middle number that matches neither. For AI traffic, the first check is time to purchase and assisted orders. For paid social, the first check is the ad's comment thread and the device that landed.
Comments Are Part of the Conversion Path
Buyers read the comment section under an ad before they click. A public answer to a real question is conversion work.
Meta lets advertisers hide or report a comment from Ads Manager, and a hidden comment stays visible to the person who posted it while disappearing for everyone else, as Meta's comment controls documentation describes. Hide spam, scams and abuse. Answer a legitimate objection. Removing a genuine complaint leaves the next shopper with the same doubt.
TikTok handles negative feedback on its own terms. Comment quality there can affect whether a campaign keeps running, not only how the brand looks. Native filters can hide or hold comments, including spam, and that path is not the same as Meta's hide or report tools.
Judge each cohort on recovered orders and assisted revenue. A blended rate that rises because you cut spend is not progress.
Diagnose a Falling Rate Stage by Stage
Take one case. Sessions hold, product views hold, add-to-cart holds, checkout starts hold, and purchases fall. Nothing upstream moved, so the break is the payment step. The first check is error rates by payment method and by device, not a homepage redesign.
Stop at the first stage that moved. Ignore every downstream rate until that stage recovers, because a payment failure suppresses the purchase rate and the revenue figure simultaneously and tells you nothing extra.
Funnel stage | Signal | First check |
|---|---|---|
Landing page view | Bounce rate rises on mobile | Page speed and match between ad and page |
Product view | Few product views per session | Product images, page load and product information |
Add to cart | Low cart rate per product view | Price display, shipping costs and offer clarity |
Checkout start | Drop after cart | Form fields, shipping and tax visibility |
Purchase | Drop at payment step | Payment errors by method and device |
Treat Speed as the Landing and Product Check
A Google-cited SOASTA study reported that mobile pages loading one second faster achieved up to a 27% increase in conversion rates, as noted in Baymard's cart abandonment analysis. Read that as an upper bound from one study, not a promise for your store. The size of your gain depends on your baseline speed, the page type, your device mix and how warm the traffic is.
Slower pages delay product comprehension and checkout progression, and they raise the odds the visitor leaves before the page is usable. Measure real-user Largest Contentful Paint, Interaction to Next Paint and layout stability alongside the same funnel events you use for conversion, so a speed regression shows up next to the drop it caused.
Fix in this order:
Compress images and serve modern formats.
Remove render-blocking scripts from the critical path.
Cut third-party tags you no longer use.
Cache static assets at the edge.
Test the ad landing page on a typical mobile network, not office wifi.
Then judge the change on that stage's progression. If checkout starts rise and purchases stay flat, you fixed one thing and exposed another, and recovering the carts you've already earned is the next lever rather than a sitewide rate target.
FAQ About Ecommerce Conversion Rate
Should refunded orders stay in the numerator
Keep the original purchase in the rate for the period it happened, and track refund rate as its own metric. Pulling refunds back into an earlier week rewrites history and makes every trend line unreliable.
How is revenue per session different
Conversion rate is orders divided by sessions. Revenue per session multiplies that rate by average order value, so a lower rate can still win. 3.0 percent at a $40 order is $1.20 per session. 2.0 percent at a $90 order is $1.80.
Does an ad comment count as a visit
No. A comment is engagement on the ad, not a session, until the person reaches your site. Read it as a pre-click signal about how the offer lands, and keep it out of the denominator.
When is a dip just noise
A week with few orders lets a handful of sales swing the rate. Compare channels on the monthly cohort, and treat a one-week swing as a prompt to open the stage table rather than a reason to change strategy.
Other Useful Guides
Conversational AI for ecommerce covers how answering buying questions on site changes progression at the product and checkout stages.
Get Started Today
Recalculate your rate for the last 30 days with the formula above, then build the stage table with sessions, product views, add-to-cart, checkout starts and purchases. The weakest stage will be obvious within ten minutes, and that's the only place worth spending this quarter's optimisation budget.
If shoppers stall at the payment and form stage, Exerta offers abandoned cart recovery by AI phone call and SMS, with the approach chosen for the cart. If visitors reach the page with unanswered product questions, an AI sales agent by voice or text can help once the page loads. Results vary with ad spend, engagement volume, offer and configuration, and Exerta doesn't guarantee that any customer will achieve similar results. If you're also producing the landing pages that feed this funnel, turning product URLs into advertorials is one way to raise the quality of the click before it arrives. When you're ready to review your own numbers, book a demo and bring the stage table with you.
Exerta helps ad-driven brands fix the stages where revenue leaks, through AI agents that answer buyers, follow up on stalled carts and moderate the comment sections shoppers read before they click. Bring your funnel numbers and we'll walk through where your rate is losing orders. Start at Exerta and see which stage is worth fixing first.


