Ecommerce and retail statistics with the source on every figure

What we mean by ecommerce statistics

Most published ecommerce statistics cannot be traced to anyone. A number appears in one article, gets quoted by a second, and within a year it is being cited by people who have never seen its origin — often long after the study behind it, if there ever was one, stopped being true. That is the problem this site exists to avoid.

Every figure we publish names the organisation that produced it, the period it describes, and the date we last checked the source still said what we quoted. Where we cannot establish an origin, we leave the number out. Several of the most widely repeated figures in retail and consumer research fail that test, and we say so on the pages where you would expect to find them.

The ground truth, and what it does not cover

The reliable baseline for US online retail comes from the Census Bureau, which reported e-commerce at $326.7 billion in the first quarter of 2026 — 16.9% of all retail sales. That figure is collected under a statistical mandate, revised openly, and comparable across quarters. It is also the only series here long enough to read trends from directly, which is why we anchor scale comparisons to it rather than to platform announcements.

Almost everything else in this field is softer. Platform-level data is voluntary and selective: companies publish reach figures readily and engagement figures rarely. Survey data measures what people say rather than what they do. Vendor research is commissioned by parties with an interest in the answer. None of that makes such data useless, but it does mean the provenance has to travel with the number, which is why our ecommerce statistics carry their source in the page itself rather than in a footnote.

What we cover

The work concentrates on four areas where consumer data and trends are actually decision-relevant:

• Commerce mechanics — what happens on a product page, why purchases fail, and what the resulting returns cost. Our product photography statistics report covers the gap between what shoppers use images for and what sites provide.
• Platforms — where selling actually happens and how the economics distribute. Our TikTok shopping statistics report sets the platform's growth against the fact that more than half its storefronts record no sales at all.
• Consumer behaviour — how people find, judge and choose products, including channels that never delivered on their promise. Our Amazon Alexa statistics report separates a very large installed base from an unmeasured commercial one.
• Retail and retention — what it costs to bring a customer back rather than find a new one. Our Starbucks customer loyalty statistics report derives the effective reward rate at each tier from the published mechanics.

How to judge any statistic you find here or elsewhere

Three questions separate retail and consumer data you can rely on from data that merely circulates. They apply to our pages as much as to anyone's.

Who was measured, and how many? A figure with a stated sample and method can be weighed. One attributed to an unnamed "study" cannot, however plausible it sounds.

Is it behaviour or opinion? Observed behaviour in testing and stated preference in a survey are different classes of evidence. A great many ecommerce statistics are the second presented as the first, because preferences are far cheaper to collect.

Who benefits from the answer? Commissioned research is not disqualified, but it belongs labelled. Where we use it, we say who paid for it and weight it accordingly.

Why the sourcing rule is enforced, not just stated

A figure is publishable here when it traces to the organisation that produced it — a regulatory filing, a government release, an industry body, or peer-reviewed work. That rule is built into the publishing system rather than left to editorial discipline: each figure is bound to an entry in its page's source list, and a figure whose source does not resolve fails the build. An unsourced number cannot reach the site by oversight.

Corrections work the same way. When a figure turns out to be wrong or its source is withdrawn, we change the page and note what changed, and every page carries a visible revision date. We would rather publish a correction than leave a number standing that someone else has already quoted. Read the full method, including what we refuse to publish and why.

The result is a smaller body of ecommerce and retail statistics than a site willing to repeat anything would carry, and a more useful one. If you are putting a number in front of someone who may check it, consumer data and trends are only worth quoting when the trail back to the original publisher is still intact.