Executive Summary
US manufacturing and wholesale look highly digital in the official statistics, and the UK’s last measured year put EDI level with the web. These statistics measure electronic ordering, not modern digital-commerce maturity.
This page reads the primary sources directly: US Census Bureau data through the February 2026 release of the 2023 Annual Integrated Economic Survey, the UK Office for National Statistics eCommerce series, and Eurostat’s EU survey. Every figure links its source and names its vintage, and where the government does not publish a number, the page says so instead of estimating.
-
Manufacturing and wholesale report high eCommerce shares under the Census definition. These are sector-specific measures, not one national B2B digital-share figure.
-
The definition is the finding. Census counts any order negotiated over EDI, extranet, or email as eCommerce. Large parts of B2B transaction processing went digital before the web.
-
The last official US channel split is a 2000 benchmark. EDI carried 88 percent of merchant wholesalers’ eCommerce sales that year. The UK last reported both channels for 2019, with nearly equal values; neither country has published the split since.
-
A current cross-sector view returned in February 2026. The first main Annual Integrated Economic Survey release restored a cross-sector eCommerce view through 2023. It still does not split orders by channel.
-
Measure your own channel mix. Split revenue by EDI or API, punchout, authenticated portal, open web, email and phone. Then measure new-customer share, exception rate and cost to serve for each channel.
There is no single official number for US B2B digital commerce
There is no single official percentage for all US B2B commerce. The Census Bureau reports sector-level eCommerce shares, and those shares are much higher in manufacturing and wholesale than in retail because the definition includes EDI, email and extranets as well as the web. The sector measures rest on different economic activities and denominators: value of shipments for manufacturing, sales for wholesale and retail, revenues for services. Census itself cautions that these differ conceptually and should be combined carefully, and its historical estimate that 94 percent of eCommerce was B2B relied on simplifying assumptions rather than direct measurement.
There is also substantial B2B activity outside manufacturing and merchant wholesale, and some transactions inside those sectors would not fit most readers’ mental model of independent B2B commerce, such as shipments between plants of the same company. So the honest form of the headline question is sector by sector, and that is how this page answers it.
What the latest US data shows, sector by sector
The current baseline arrived in February 2026. The Census Bureau published the first main release of the Annual Integrated Economic Survey, which replaced seven separate annual business surveys starting with the 2023 survey year. Its data products include a cross-sector eCommerce table for employer firms. For the first time in years, the numbers sit in one place: one release, one reference year, one survey universe, running through 2023. The sector measures underneath still do not match. But you are no longer stitching them together from seven separate surveys. These final estimates replace the preliminary figures released in July 2025.
Census eCommerce share by sector, 2023, one table, one year.
shipments, 2023
manufacturers' sales branches
and offices, eCommerce share of sales, 2023
Source: US Census Bureau, 2023 Annual Integrated Economic Survey, table AIES00ECOM, main release February 26, 2026, accessed August 2, 2026. eCommerce as a percent of total sales, value of shipments, or revenue.
For context on where the series has been: the last E-Stats report, covering 2019, put eCommerce at 67.8 percent of manufacturing shipments, 33.3 percent of merchant wholesale sales including manufacturers’ sales branches, and 10.7 percent of retail sales. The ranking, manufacturing ahead of wholesale ahead of retail, has held for as long as the government has measured it. Against 2019, the 2023 figures show manufacturing essentially flat, wholesale up about two and a half points, and retail up nearly five, the sharpest rise of the three. That order is stable. What it does not explain is why manufacturing and wholesale sit so far ahead of retail, and the answer is in what the surveys count as eCommerce.
Historical source: US Census Bureau, 2019 E-Stats release, August 2021.
Why manufacturing and wholesale look more digital than retail
Because of what the word eCommerce means in these surveys. Census counts a transaction as eCommerce when the order is placed, or the price and terms are negotiated, over the internet, an extranet, an Electronic Data Interchange network, electronic mail, or another comparable online system. Payment does not have to be made online. Nothing about it has to touch a website.
EDI is the load-bearing clause. Electronic Data Interchange is a standardized method through which large trading partners have exchanged purchase orders, invoices and shipping notices machine to machine for decades. When a vehicle assembler’s ERP sends an X12 850 purchase order to a brake supplier’s EDI gateway, that shipment is eCommerce in the federal statistics; transportation equipment led all manufacturing subsectors at 80.7 percent of shipments in the 2023 data. Nobody browsed a page and no product image was rendered; two systems configured to talk to each other years ago exchanged a structured message.
A company can report a high digital share while new-customer discovery and self-service ordering remain weak. In that case, the number is measuring transaction automation, not digital-commerce maturity.
That is the distinction the official statistics cannot make for you. The commercially relevant line does not run between digital and nondigital. It runs between electronically transacted revenue and discoverable, self-service digital commerce. The Census definition spans both, but the published total does not show how much sits in each. Setting each ordering channel side by side shows exactly where it falls short of open, self-service commerce.
What Census counts that leaders would not
Lay the ordering channels against what each one proves and the definitional gap becomes a working diagnostic. Every row below counts as Census eCommerce except phone, although they differ materially in what they prove about discovery, self-service and automation.
What Census counts as eCommerce versus what leaders mean by digital commerce.
| Channel | Counts as Census eCommerce | Supports open discovery | Supports self-service evaluation | Requires an established relationship | Suitable for agent access |
|---|---|---|---|---|---|
| EDI | Yes | No | No | Yes | With integration |
| Punchout or API | Yes | No | Within the buyer's system | Yes | With integration |
| Authenticated portal | Yes | No | After authentication | Yes | After authentication |
| Public web store | Yes | Yes | Yes | No | Yes, if data is structured |
| Email order | Yes | No | No | Usually | No |
| Phone order | No | No | No | Usually | No |
The table shows which channels can carry open discovery and which cannot. The next question is how much B2B revenue actually runs on each, and that is where the public record goes quiet.
Where the EDI-versus-web split stands in the US, UK and EU
The Census Bureau does not publish a current national EDI-versus-web split, and we found no authoritative US public replacement. The last official Census breakout was published for 2000, and it is worth keeping precisely because it is historical. That year, EDI carried $188 billion of merchant wholesalers’ eCommerce sales, 88 percent of the total. Manufacturing plants that primarily used EDI to accept online orders accounted for roughly two thirds of eCommerce shipments among responding plants, while plants primarily using internet networks accounted for 5 percent. Treat those figures as a 2000 benchmark, not an estimate of today’s mix. The US record stops there. Britain’s runs almost two decades further.
Source: US Census Bureau, E-Stats 2000 edition, the last release with a mode-of-ordering breakout.
The UK kept counting until 2019
The Office for National Statistics ran an eCommerce survey on the same definition, splitting sales between EDI and websites. Its last release covers 2019: £668.9 billion in eCommerce sales from UK non-financial businesses with 10 or more employees. The published point estimates were almost identical: £334.0 billion through EDI and £334.9 billion through websites, reached from very different directions, with website sales having more than trebled since 2009 while EDI sales grew by about a quarter. ONS cautioned that both estimates carried wider confidence intervals than in 2018, so the useful finding is parity in the reported values rather than a statistically tested crossover. The 2019 bulletin remains the last ONS release to report EDI and website sales together. A later Digital Economy Survey, published in April 2023, gave website sales for 2021 but no EDI estimate. ONS has since paused that survey to rebuild it over data-quality concerns, so the channel split was never resumed.
The last UK year with both channels reported: values nearly equal, and the split not continued since.
Source: Office for National Statistics, eCommerce and ICT activity, UK: 2019, published February 2021, the latest release in the series. Figures as revised in that bulletin.
Web stores, portals, punchout catalogs and marketplaces have grown since 2000. EDI remains widely used, but the national division between those channels is no longer published in the US or the UK. The EU still measures it: Eurostat’s 2024 data shows EU enterprises taking more turnover through EDI-type sales, at 11.07 percent, than through websites and apps, at 8.39 percent, with the EDI share highest in manufacturing. Our upcoming companion page on machine-readable product data carries that chart and its sources. The public statistics stop at the national level. Your own channel mix is the one measurement you fully control.
What B2B leaders should measure internally
The number the government does not produce is the one you can. Split your own revenue by ordering channel: EDI or API, punchout, authenticated portal, open web, email and phone. Then measure three things per channel: new-customer share, exception rate, and cost to serve. Automated repeat trade proves known counterparties can transact efficiently. An authenticated portal proves existing customers can search, price and order. Open digital commerce provides the clearest direct evidence that unfamiliar buyers can discover and evaluate you without an existing account or human intermediary, and the exception rate tells you how much of your digital revenue still depends on manual interpretation. Every channel in that list assumes the buyer already knows you. The next shift in B2B does not.
What agent-readable commerce adds to old rails
EDI proves that B2B companies have automated repeat transactions between known counterparties for decades. Agentic commerce introduces a harder problem: discovery and product understanding before the trading relationship exists.
An EDI purchase order usually assumes that both parties already know the product identifier, price structure and commercial relationship. EDI messages can carry identifiers and descriptions; what the standard was never designed to be is an open discovery and product-understanding layer. An AI procurement journey may begin with an open-ended requirement. It needs to find candidate products, compare specifications, resolve compatibility and determine which supplier can fulfill the request, all before the order ever reaches an API, portal or EDI workflow. That front end runs on structured, machine-readable product data, which is a different asset from a transaction pipe, and many established B2B businesses have invested in transaction rails for decades and in open, machine-readable discovery for much less time.
AI adoption is now material, particularly among larger firms: the Census Bureau’s Business Trends and Outlook Survey shows overall use holding between 17 and 20 percent of US firms from December 2025 to May 2026, and 37 percent among firms with 250 or more employees. Those figures cover US businesses generally rather than B2B companies specifically, and Census does not measure the overlap between AI adoption and EDI operation directly. The relevant point for this page is architectural: many companies are adding probabilistic, language-driven systems on top of transaction flows designed for known products and established trading relationships. That gap between old rails and new discovery is not abstract. It surfaces in the systems companies run day to day.
How Hancocks and Wemoto turned electronic revenue into self-service commerce
The distinction matters in real platforms. Hancocks, a UK bulk confectionery wholesaler trading since 1962, already had substantial digital trading activity when Net Solutions rebuilt its platform on headless Adobe Commerce with Contentful; its legacy architecture restricted concurrency, content operations and self-service growth. The modernization work was not about making the business digital for the first time. It was about turning existing electronic revenue into a scalable, discoverable commerce operation, without pausing live B2B trading while it happened. The Wemoto consolidation, eight legacy systems into one multi-tenant platform with a UK trade portal and automated cross-franchise ordering, is the same distinction at a different scale, and both case studies carry their published results.
The B2B platforms that win the next five years will not be the ones with the highest eCommerce share. They will be the ones a stranger, or a stranger’s agent, can read, price, and buy from without picking up the phone. The statistics already say most of the market is digital. Very little of it is legible.
Method note. All US figures come from Census Bureau publications, linked at first use with vintages stated per figure; the EU split comes from Eurostat. Sector measures are never combined into a single national B2B figure, because the underlying economic measures differ. Where no current figure exists, the page says so rather than estimating. This page is rechecked against each AIES eCommerce release; superseded values remain visible so direction can be inspected. Figures last verified August 1, 2026.
Common questions
There is no single official percentage for all US B2B commerce. The Census Bureau reports sector-level eCommerce shares, and those shares are much higher in manufacturing and wholesale than in retail because the definition includes EDI, email and extranets as well as the web. The sectors use different economic measures, so they do not sum to one national B2B figure.
Yes. Census counts any order where price and terms are negotiated over the internet, an extranet, an EDI network, email, or another online system. Payment does not have to be made online. A purchase order moving through a decades-old EDI connection counts the same as a web checkout.
The Census Bureau does not publish a current national EDI-versus-web split, and we found no authoritative US public replacement. The last official Census breakout, for 2000, put EDI at 88 percent of merchant wholesalers’ eCommerce sales. The UK’s last reported split, for 2019, showed nearly equal values at roughly £334 billion each way, and the split has not been continued since. The EU still measures the split: Eurostat’s 2024 data shows EDI-type sales carrying more enterprise turnover than web sales.
Yes. In February 2026 the Census Bureau published the first main release of the Annual Integrated Economic Survey, including a cross-sector eCommerce table for employer firms with data through 2023. What it does not publish is a breakdown of how those electronic orders were placed: EDI, web storefront, portal, punchout or another channel.
eCommerce share measures how much revenue was ordered or negotiated electronically. Digital-commerce maturity asks a different question: whether customers can discover products, evaluate them, receive account-specific prices, complete orders without manual intervention and obtain consistent service across channels. A company can score highly on the first measure and poorly on the second.
Which pipe does your digital
revenue run on?
Bring your order-channel data and platform architecture. We will help separate automated repeat trade from discoverable self-service commerce and identify where modernization can reduce cost to serve or open new demand.
