Most benchmark posts quote one median with no range, no sector and no source. Here are eight numbers with quartiles, the confidence behind each, and why you can't multiply them together.
Every benchmark post you've read gives you one number. "The average ecommerce conversion rate is 2.5%." No range, no sector, no device split, and usually no source you can check.
That number is useless in both directions. If you're at 1.8% you don't know whether you have a problem or a category. If you're at 3%, you don't know whether you're good or whether the average was dragged down by stores that don't matter.
So here are eight, with quartiles, and with an honest note on how much each one is worth.
The eight
1. Session conversion rate
- All — p25 0.9-1.2%, median 1.4-2.3%, p75 3.2%, p90 4.7%, confidence M
- Desktop — p25 1.1%, median 1.9%, p75 3.5%, p90 6.5%, confidence H
- Mobile — p25 0.7%, median 1.2%, p75 2.0%, p90 3.9%, confidence H
Read the device split before the blended number. Most of your traffic is mobile, so your blended rate is mostly a mobile rate with a desktop halo. Teams optimize on the desktop view because that's the screen in front of them, and then wonder why the number doesn't move.
By sector the median ranges from about 1.43% in automotive to 3.00% in food and beverage. If you sell furniture and you're comparing yourself to supplements, you're going to reach a wrong conclusion.
2. Add-to-cart rate
Median around 6.07% across all sectors. By category it swings enormously: roughly 1.72% in luxury and jewelry, around 17.02% in beauty and personal care.
That spread is the most useful thing on this page. Add-to-cart tracks how considered the purchase is. A low rate in jewelry is normal. The same rate in beauty means something is broken.
3. Cart-to-purchase rate
Food and beverage runs near 41%, pets near 44%, apparel near 18%, automotive near 13%.
Pair it with add-to-cart. A healthy add-to-cart and a poor cart-to-purchase points at cost surprise or checkout friction. The reverse points at the product page.
4. Cart abandonment
Median 70-74%, best-in-class near 52.8%. Confidence: H, from a fifty-study meta-analysis.
Luxury and jewelry sit above 81%. Pets near 53%. This has barely moved in a decade of better tooling, which should temper expectations about what any single fix will do.
5. Average order value
- All stores — p25 $45, median $70-85, p75 $192, p90 $311
Never read AOV alone. It rises when you lose small orders, so a store getting worse at converting budget buyers shows a healthier AOV and a worse revenue line.
6. Gross and contribution margin
- Gross margin — p25 46%, median 57%, p75 64%, p90 70%+
- Contribution margin — p25 3%, median 15-20%, p75 25%, p90 56%
Contribution margin is the one that governs decisions, and it's the one nobody quotes. It's what's left after product, shipping, fees and acquisition. At the median you keep 15 to 20 cents on the dollar, which is why a 20% discount code isn't a haircut. It's the whole thing.
7. Repeat purchase rate
12-15% at p25, 20-28% median, 35% at p75, 45% at p90. Confidence: L, directional only.
This decides whether discounting a first order is acquisition or just a smaller sale. Most stores have never calculated it and use a number they heard.
8. Return rate
Roughly 16.9-19.3% overall, and contested. Confidence: L.
Heavily category-dependent. Apparel runs far above the average, most consumables far below. Returns are a margin line that rarely appears in a conversion discussion, which is how a successful test can lose money.
Two things that will trip you up
Medians don't compose. You can't take median conversion rate, multiply by median AOV, and get median revenue per session. The store at median conversion isn't the same store as the one at median AOV, and they're often opposites: high-consideration stores convert badly and sell expensively. Multiply the two and you invent a business that doesn't exist.
Conversion rate is roughly flat with revenue. This one surprises people. Scaling to $50M doesn't improve your conversion rate. What moves with scale is add-to-cart rate and customer acquisition cost. So you can't grow your way out of a conversion problem, and a store doing $200k a month should compare itself to the same benchmark as one doing $20m.
How to use these
Pick your sector row, not the all-sector row. Then split by device, then find your quartile.
If you're below p25 on a metric, that's a defect and it's usually specific. If you're between p25 and median, that's ordinary and improvable. Above p75, leave it alone and go work on something else, because the remaining gain is smaller than the gain available elsewhere.
What benchmarks can't do is tell you why. Knowing your mobile conversion sits at 0.8% against a 1.2% median tells you there's a problem worth about a third of your mobile revenue. It doesn't tell you that your price sits below a gallery that eats the first screen on a phone.
That's the gap our audit fills: 500 checks across the store, on mobile and desktop, every page, each scored against a bar rather than an opinion. Then we build the change and adjudicate it against real revenue, both halves, so the improvement arrives as a number you can put next to these ones.
You can get partway there with an afternoon and your own phone. Most founders find two or three real problems in twenty minutes. The catch is that the store changes every week and the afternoon happens once.
Sources
Every figure above comes from published research, graded by confidence:
- Session CVR: Littledata (~2,800 Shopify sites), Contentsquare Digital Experience Benchmark 2026 (99B sessions, 6,500+ sites), Polar Analytics (4,000+ Shopify brands)
- Add-to-cart and cart-to-purchase: Dynamic Yield (200M+ sessions), Polar Analytics
- Cart abandonment: Baymard Institute (50-study meta-average), eMarketer / Dynamic Yield
- AOV: Polar Analytics, dtcpages aggregation
- Margins: LevelCFO Ecommerce Benchmarks 2026 with SEC 10-K filings from 11 public DTC brands
- Repeat purchase and returns: Eightx (aggregating Rivo, Yotpo, MageLoyalty), Richpanel
Where a number is graded L, treat it as a direction rather than a target.
