Reference data

E-commerce statistics

How much of US retail is actually transacted online, who the sellers are, and why five commonly interchanged terms — e-commerce, online retail, electronic shopping, nonstore retail and digital commerce — measure five different things. Every figure traces to a federal dataset, and every calculation Fundur made is shown in full.

E-commerce sales figures are the Census Bureau’s Quarterly Retail E-Commerce Sales release for 2026 Q2, last revised 18 August 2026. Establishment and payroll figures are County Business Patterns, reference year 2023. Firm counts are Statistics of US Businesses 2022. Returns figures are the National Retail Federation’s 2025 Retail Returns Landscape. SBA figures are computed by Fundur from SBA’s 7(a) FOIA loan-level file, snapshot 30 June 2026. Last reviewed 28 August 2026.

17.1%of US retail sales were e-commerce in 2026 Q2, seasonally adjusted

$1.234Tof US retail e-commerce in calendar 2025

42.1%of that e-commerce happens at retailers not classified as nonstore retailers

54,707electronic shopping and mail-order establishments in the United States

Definitions

Five terms, five different measurements

Most disagreement between published e-commerce numbers is not a data disagreement. It is that the writers were counting different things. This is the part of the page worth reading before the numbers.

What each term measures in official US statistics
Term	What it actually is	Official source	Latest measured value
E-commerce	A transaction method: any retail sale where the order was placed, or price and terms agreed, over an online system. Applies to any retailer, store or otherwise.	Census Quarterly Retail E-Commerce Sales	$1.234tn in calendar 2025
Nonstore retailers	An industry: NAICS 454, retailers whose primary channel is not a physical store — electronic shopping, mail order, vending, direct selling, fuel dealers.	Census Monthly Retail Trade Survey	$1.519tn of sales in calendar 2025
Electronic shopping and mail-order houses	A narrower industry: NAICS 4541, one subsector inside nonstore retail. Includes catalogue and telephone order, not only online.	Census MRTS / County Business Patterns	$1.431tn of sales; 54,707 establishments
Online retail	No official definition. Used interchangeably for all three of the above, which is why figures labelled this way rarely reconcile.	—	—
Digital commerce	No official definition. Usually broader than retail e-commerce, often folding in services, B2B or travel that Census retail figures exclude entirely.	—	—

The single most common error. Treating “nonstore retailers” or “electronic shopping and mail-order houses” as a headcount of e-commerce businesses, and treating e-commerce sales as their revenue. They are an industry classification and a transaction method respectively. Neither contains the other, and the two overlap only partially — as the next two sections show, using nothing but Census’s own arithmetic.

The measured share

E-commerce is 17.1% of US retail sales

The Census Bureau has measured retail e-commerce quarterly since 1999. This is the authoritative series, and the one every credible share figure ultimately traces back to.

US retail sales and e-commerce sales, not seasonally adjusted, $ millions
Quarter	Total retail	E-commerce	E-commerce share
2026 Q2 (p)	2,011,977	329,530	16.4%
2026 Q1 (r)	1,804,536	301,242	16.7%
2025 Q4	1,999,440	365,160	18.3%
2025 Q3	1,896,573	299,717	15.8%
2025 Q2	1,887,093	293,140	15.5%
2025 Q1	1,735,586	275,715	15.9%
Calendar 2025	7,518,692	1,233,732	16.41%

17.1%

e-commerce share of total US retail sales in 2026 Q2 on the seasonally adjusted series, against 16.4% on the unadjusted series for the same quarter. Both are published; they are not interchangeable, and the adjusted figure is the one Census leads with.Census Bureau, Quarterly Retail E-Commerce Sales, Table 1 · 2026 Q2, revised 18 August 2026

18.3%

was the unadjusted share in 2025 Q4 — the annual high. Unadjusted e-commerce share peaks every fourth quarter and drops back in the first, which is why a single quarter quoted without its adjustment basis can be misleading by nearly three points.Census Bureau, Quarterly Retail E-Commerce Sales

±2.2%

coefficient of variation on the quarterly e-commerce estimate. It is a survey estimate with published sampling error, not a census count, and Census publishes the measures of variability alongside it.Census Bureau, Quarterly Retail E-Commerce Sales, Table 2

Original analysis

42% of US retail e-commerce happens outside the nonstore sector

Census publishes e-commerce sales broken out by kind of business. Set the nonstore retailers line against the total and the answer is stable across every quarter currently published.

Nonstore retailers’ e-commerce as a share of all US retail e-commerce, not adjusted, $ millions. Share columns computed by Fundur.
Quarter	All retail e-commerce	Of which nonstore retailers	Nonstore share	Not nonstore
2026 Q2 (p)	329,530	190,826	57.9%	42.1%
2026 Q1 (r)	301,242	173,661	57.6%	42.4%
2025 Q4	365,160	209,707	57.4%	42.6%
2025 Q3	299,717	174,608	58.3%	41.7%
2025 Q2	293,140	171,507	58.5%	41.5%

In every quarter Census currently publishes at this level of detail, between 41.5% and 42.6% of US retail e-commerce sales were made by retailers that are not classified as nonstore retailers. That is grocers, clothing retailers, general merchandise chains, building suppliers and car parts sellers — store-based businesses selling online.

The reverse also holds and is quoted even less often: nonstore retailers are not wholly online either. In 2026 Q2 the sector sold $290,330 million in total, of which $190,826 million was e-commerce — 65.7%. Roughly a third of nonstore retail is still mail order, telephone order, vending and direct selling.

The $197.6 billion arithmetic. In calendar 2025, electronic shopping and mail-order houses (NAICS 4541) recorded $1.431 trillion of sales, while all US retail e-commerce was $1.234 trillion. One industry outsold the entire transaction method by $197.6 billion — from the same Census survey, the same period, the same universe. There is nothing wrong with either number. An industry total includes that industry’s non-online sales, and a transaction-method total includes everybody else’s online sales. If a published figure treats the two as the same thing, this is the size of the error.

Measurement breaks

Two breaks that make long e-commerce time series unsafe

Both are documented by the agencies themselves, and both are routinely crossed without comment in published charts.

Apr 2025

The universe changed. In Census’s own words, quarterly e-commerce estimates “include data only for businesses with paid employees. Prior to the benchmark report released in April 2025, the Quarterly E-commerce estimates included nonemployers.” A chart running through that point is comparing two different populations.Census Bureau, Quarterly Retail E-Commerce Sales, Table 1, footnote 1

The industry code was retired. In the 2022 revision of the North American Industry Classification System, 454110 Electronic Shopping and Mail-Order Houses was split across 42 different 2022 retail industries by merchandise line, and no 2022 code begins with 454 at all. There is no longer any single industry code that means “e-commerce business”.Counted by Fundur from the US Census Bureau 2017-to-2022 NAICS concordance

The second break is visible in federal lending data. SBA records a NAICS code on every 7(a) approval, assigned at approval time, so the file shows the reclassification working through in real time as lenders adopted the new codes:

SBA 7(a) approvals by NAICS code and fiscal year, cancelled approvals excluded. Computed by Fundur from SBA loan-level data.
NAICS	FY2020	FY2021	FY2022	FY2023	FY2024	FY2025
454110 Electronic shopping and mail-order houses	291	416	293	48	18	10
459999 All other miscellaneous retailers	0	1	104	292	377	358
455219 All other general merchandise retailers	0	2	97	261	280	289
458110 Clothing and accessories retailers	0	0	156	393	625	546

Approvals coded to electronic shopping fell from 416 to 10 in four years. Retail borrowing did not fall — retail 7(a) approvals rose from 4,760 in FY2020 to 7,241 in FY2025. The businesses moved code. Any count of “e-commerce businesses” built from an industry classifier and run across FY2022 will show a collapse that did not happen.

The sellers

Who the electronic shopping businesses actually are

Taken as an industry rather than as a proxy for e-commerce, NAICS 4541 is small, unusually micro-heavy at the firm level, and unusually well paid at the employee level.

54,707

electronic shopping and mail-order establishments, employing 881,019 people — 5.2% of retail establishments and 5.5% of retail employment.Census Bureau, County Business Patterns 2023, NAICS 45411

$58,522

average annual payroll per employee, against $37,746 across retail as a whole. On a headcount basis these are the best-paid jobs in the retail sector — warehousing, merchandising and technology rather than shop floor.Fundur calculation from County Business Patterns 2023

74.2%

of electronic shopping firms have fewer than five employees, against 59.9% across retail and 63.0% across all US employer firms. It is the most micro-heavy corner of retail, and the establishment average of 16.1 employees hides that entirely.Fundur calculation from Statistics of US Businesses 2022

19.3%

of online sales were expected to be returned in 2025, against 15.8% across all retail channels on the same survey.National Retail Federation and Happy Returns, 2025 Retail Returns Landscape · 15 October 2025

Financing

Why SBA data understates e-commerce borrowing

The obvious way to measure online-seller borrowing is to count SBA approvals under the electronic shopping code. The measurement breaks above are exactly why that number should not be used on its own.

1,076

SBA 7(a) approvals were coded to electronic shopping and mail-order houses across FY2020–FY2025 — 0.36% of all 7(a) approvals, totalling $445 million. The code was retired part-way through the window, so this is a floor, not a count of online-seller borrowing.Fundur calculation from SBA 7(a) FOIA loan-level data

$117,500

median approval under that code, against $250,000 across retail and $198,300 across all 7(a) — the smallest median of any industry group on these three pages. The mean is $413,541, 3.52× the median: the widest mean-to-median gap here.Fundur calculation from SBA 7(a) FOIA loan-level data

10.93%

of approvals under this code were later cancelled and never disbursed, against 12.93% across all 7(a).Fundur calculation from SBA 7(a) FOIA loan-level data

Method

How these numbers were produced

E-commerce sales figures are read from the Census Bureau’s Quarterly Retail E-Commerce Sales release for 2026 Q2, last revised 18 August 2026 — Table 1 for the totals and adjusted share, Table 2 for sampling variability, and the supplemental tables for the kind-of-business detail. The calendar 2025 total is the four not-adjusted quarterly e-commerce estimates summed; the calendar 2025 retail total is the four matching not-adjusted total estimates summed, and it equals $7,518,692 million, which is exactly the calendar 2025 total in the separate Monthly Retail Trade Survey time series. That equality is the check that the two series share a universe.

The share columns in the nonstore table are ours: nonstore-retailer e-commerce divided by total e-commerce, quarter by quarter, from the supplemental table. No modelling, no adjustment, no interpolation. The $197.6 billion figure is one subtraction: NAICS 4541 calendar 2025 sales of $1,431,304 million from the Monthly Retail Trade Survey, less calendar 2025 total e-commerce of $1,233,732 million.

Establishment, employment and payroll figures come from the County Business Patterns 2023 US summary file, rows where legal form of organisation is set to all. Firm-size figures come from Statistics of US Businesses 2022, the most recent year published. Both cover establishments and firms with paid employees only.

SBA figures are computed by Fundur from SBA’s 7(a) and 504 FOIA loan-level file, 7(a) extract, snapshot dated 30 June 2026 (SHA-256 beginning 6c1e9132, 181,130,871 bytes). From 388,338 parsed rows we keep complete fiscal years 2020 to 2025, gross approval above zero and a valid approval date, then exclude the 12.93% of approvals recorded as cancelled and never disbursed, leaving 302,593 approvals. Industry assignment uses the NaicsCode field as recorded on each approval, which is why the code-migration table can be read at all. The file contains 368 exact duplicate rows in this window, 0.11%; removing them does not change the $117,500 median or any percentage on this page. No rate filter is applied because this page publishes no interest-rate figure.

What these numbers cannot tell you. Census retail e-commerce covers retail trade only. It excludes services, travel, ticketing, business-to-business trade and food service delivery, all of which are commonly included in commercial “digital commerce” totals — which is the main reason private estimates run far above Census. Since the April 2025 benchmark it also excludes businesses with no employees, so marketplace sellers below the payroll threshold are outside it entirely. The quarterly estimate is a survey with a published coefficient of variation of roughly 2.2%, not a count. County Business Patterns and Statistics of US Businesses cover employers only and come from two different reference years. The SBA file records approvals rather than applications and covers only the 7(a) programme; the code-migration table shows a classification change, which is not evidence about how many online sellers borrowed.

Reuse and citation

You are welcome to reproduce any figure or table on this page in noncommercial reporting, research or educational material, with attribution and a link back to this page. No permission request is needed. Where a figure is a Fundur calculation rather than a published statistic it is labelled as one above, and we would rather you carried that label through.

If a figure here contradicts one you have published, we would rather you checked our method than took it on trust — it is set out in full above, and every input file is linked.

Suggested citation: Fundur, “E-commerce Statistics,” last reviewed 28 August 2026. https://fundur.com/resources/ecommerce-industry-statistics/

Sources

Every source on this page

US Census Bureau — Quarterly Retail E-Commerce Sales: Table 1, Table 2, supplemental tables · 2026 Q2, last revised 18 August 2026
US Census Bureau — Monthly Retail Trade Survey sales time series, 1992 to present · calendar 2025, not adjusted
US Census Bureau — County Business Patterns, 2023 US summary data file · establishments, employment, annual payroll
US Census Bureau — Statistics of US Businesses, 2022 US and state detailed employment sizes · firms by employment size
US Census Bureau — North American Industry Classification System and the 2017-to-2022 concordance · 2022 revision, retail sector reorganisation
National Retail Federation and Happy Returns — 2025 Retail Returns Landscape · press release 15 October 2025
US Small Business Administration — 7(a) & 504 FOIA loan-level data · 7(a) extract, snapshot 30 June 2026

Fundur is a financing marketplace, not a lender. If you sell online and are working out what your own business could borrow, our e-commerce business loans page covers the options and what lenders weigh. The wider retail picture is on our retail industry statistics page.