EDI, Older Than Your Dad And Still Runs Global Trade.

Four large hotel chains came on at the same time.

All four wanted to trade electronically.

All four used a different protocol.

One was SFTP.

One was API.

One was XML.

One was doing something I had to google.

Each had different catalogue requirements. Each had different contract pricing. Each had a different person on the other end who would go quiet for three weeks and then resurface with a spec sheet that contradicted the last one.

My actual job kept getting pushed into the margins. Instead I was project managing four parallel integration builds that each required specialist knowledge, custom mapping work, and a testing cycle that took longer than the integration itself.

That wasn't unusual.

That was a Tuesday.

At a role managing digital channels, EDI onboarding ate months of staff time on a continuous rolling basis. A new supplier or customer would come on excited to trade digitally. Then someone, somewhere would say "EDI integration" out loud and the timeline would stretch to four, sometimes six months. By the time the connection was live, the commercial moment had passed, the internal champion had moved on, and you'd absorbed the cost in staff time, consultant fees, and goodwill.

This isn't a niche problem. its how most global b2b commerce actually moves.

And almost nobody outside operations teams knows it exists, and the outsiders just harp on about ‘API’s’.

The protocol that time forgot

Electronic Data Interchange was invented in the 1960s. Companies needed to exchange structured business documents, purchase orders, invoices, shipping notices, without humans rekeying everything. Before EDI, that meant telex machines, paper forms, and an army of data entry clerks.

EDI fixed it.

For the 1960s.

By the 1980s, major retailers had made it a condition of doing business. Want to sell to Walmart?

Speak EDI.

Want to supply a car manufacturer?

Speak EDI.

The protocol spread through supply chains by pure power dynamic. Whoever was bigger mandated. Whoever wanted access complied.

The dominant standards are X12 in North America and EDIFACT internationally.

Both are structured.

Both are document based.

Both are deeply unglamorous and extraordinarily embedded (see ‘api’).

If you've never heard of them, that's because they live inside the operational plumbing of global trade.

nobody at TechCrunch writes articles about operational plumbing.

But parts are broken.

The standards defined the document structure.

They did not define the implementation.

Every trading partner implemented the standard differently.

Different field requirements.

Different code sets.

Different validation rules.

Different SFTP configs, different AS2 setups, different file naming conventions.

Built for the ERP of the demanding party.

The word "standard" in EDI is one of the great misnomers in enterprise software.

Up there with "cloud native" and "self service."

So every new trading partner connection is, functionally, a bespoke engineering project. You're not plugging into a standard. You're translating between your interpretation of the standard and theirs.

Multiply that by every supplier, every retailer, every logistics partner, every procurement system you connect to.

Then layer punchout on top of it.

Punchout arrived in the late 1990s when SAP Ariba and later Coupa became the dominant procurement platforms for large enterprises.

Punchout is the shop window, or for you d2c kids, its literally a eCommerce store, kinda.

A buyer logs into their procurement system, punches out into the supplier's catalogue, shops, and the cart flows back as a requisition.

Punchout didn't replace EDI.

It sat on top of it.

In most chains, both coexist.

A buyer punches out to browse products. The resulting purchase order comes back via EDI. The invoice goes back via EDI. The shipment notice goes back via EDI.

Punchout is the front of house.

EDI is the back of house pipe.

Neither is standardised in any way that makes connecting two trading partners a predictable exercise.

The size of the quiet beast

The EDI software market was worth $34 billion in 2024.

Projected to hit $74 billion by 2031.

$34 billion.

Bigger than the global coffee shop chain market.

Bigger than the global yoga industry.

Bigger than almost everything you'd pick in a trivia category called "industries worth more than you think."

The dominant vendors are SPS Commerce, TrueCommerce, Cleo, IBM Sterling, and OpenText.

Never heard of them?

Exactly.

This is mostly legacy businesses with managed service models.

Their revenue is built on complexity. The more bespoke each integration is, the more billable hours for setup, mapping, and ongoing compliance management. Per partner setup fees run $500 to $5,000. Per transaction fees. Monthly subscription. Years long partnerships that talk bad about the people on the other side of the fence.

A support desk handling the inevitable breaks when a trading partner changes their spec without telling anyone.

SPS Commerce is publicly listed. Market cap around $4B on $600M annual revenue.

Well run, sticky, and profitable.

Also economically incentivised to keep the problem complex.

When your business model is billable hours, the last thing you want is to make the work take less time.

The New Kids On The Block

Orderful raised 40 mil and some change, including from Andreessen Horowitz. Logistics and retail EDI with a pre-mapped network model. EDIFACT support. AI assisted mapping. Pre-mapped trading partners.

They claim 9 days to go live with common partners.

Nine days versus four months is not nothing.

They're the closest thing to a modern EDI platform.

North American retail is their home turf.

That's where the pre-mapped network is densest, that's where the trading partners are, that's where their moat is strongest.

Stedi raised $142M total.

Pivoted sharply into healthcare EDI specifically after the Change Healthcare outage in 2024.

Their insight was sharp.

Healthcare EDI has unusually consistent standards because HIPAA mandates them. More tractable than general supply chain EDI. Smart move. Not a play on the broader B2B commerce problem.

AWS entered in late 2023 with B2B Data Interchange.

A managed service for EDI transformation.

Late 2024 they added generative AI assisted mapping. Upload your EDI document and your target JSON format, AI generates the mapping.

Early users report 50% reduction in time and expertise required.

The catch?

X12 only. No EDIFACT. And it solves the transformation problem, not the connectivity or onboarding problem. You still need to establish the trading partner relationship, configure the protocols, manage the testing cycle.

Everyone is adding AI to the mapping layer.

Nobody has rebuilt the onboarding experience end to end.

The rest of the world

The US is the biggest market, the most fragmented, and the most dominated by legacy infrastructure.

North America accounts for 51% of the global EDI software market.

Every major retailer, Walmart, Target, Costco, Kroger, Amazon, mandates EDI from suppliers.

Europe is pushing hard on standardisation through ViDA, which mandates e-invoicing and digital reporting across all 27 member states by 2030.

Germany's B2B e-invoicing mandate is live for receiving from 2025. France, Poland, Romania on similar trajectories.

Peppol is the shared plumbing.

Europe's approach is structurally the most interesting. By mandating a common standard across 27 member states, regulation is doing what the EDI industry never managed voluntarily.

Forcing convergence on a single protocol.

Nothing like a bureaucratic hammer to get shit done.

China went its own way entirely.

The Golden Tax System is state controlled digital invoice infrastructure. E-fapiao became mandatory for B2B and B2G transactions nationwide in December 2024.

Every transaction submitted to the State Taxation Administration in real time. Cleared, returned, monitored at national level.

That's not EDI in the Western sense.

It's a state-mandated clearance system every business in China has to plug into.

Foreign operators need a B21 licence with foreign capital capped at 50%. Separate market, separate rulebook, separate conversation.

Then there's NZ and AU.

Not backwaters, and id say ahead of most markets on regulatory push.

Australia mandated Peppol e-invoicing for all federal government agencies in July 2022.

Over 400 government entities active.

Target of 30% of supplier invoices processed via Peppol by July 2026, full automation by December 2026.

B2B e-invoicing remains voluntary, but the Business E-Invoicing Right gives companies the legal right to require Peppol e-invoices from trading partners. Practical mandation by contract is already happening.

New Zealand is following the same trajectory.

From January 2026, government agencies must have e-invoice capability.

From January 2027, agencies must require large suppliers (revenue above NZD $33M) to submit via Peppol.

Every NZ and AU supplier wanting to sell to government, or to large enterprises mimicking government, needs a Peppol compatible pathway.

On top of existing EDI complexity.

Not replacing it.

More standards to not implement the same way.

N = M problem, solved everywhere except here

In November 2024, Anthropic released the Model Context Protocol (MCP).

By early 2026, it had 97 million monthly SDK downloads. OpenAI, Google DeepMind, Microsoft, AWS all adopted it.

The whole thing got donated to the Linux Foundation under vendor neutral governance.

What MCP solved, in one sentence?

The N = M integration problem for AI systems.

Before MCP, connecting an AI model to enterprise tools required a bespoke integration for every combination.

Ten AI apps, a hundred tools. Potentially a thousand integrations. Every new tool meant a new connector. Every new AI model meant every connector rebuilt.

MCP collapsed that to a single protocol.

Any AI model speaking MCP can connect to any tool exposing an MCP server. One time build. Universal compatibility. The network effect compounds. Every new MCP server is immediately accessible to every MCP compatible client.

EDI has the exact same N = M problem.

And it's had it for sixty years.

Every new trading partner connection requires building a bespoke translator between two implementations of a "standard" that nobody implements the same way. A thousand suppliers, five hundred retailers, potentially hundreds of thousands of integration permutations. The problem was never the document format. X12 and EDIFACT are well understood. The problem was the absence of a universal abstraction layer that handled translation, protocol negotiation, and onboarding without requiring a custom project every time.

MCP solved this for AI tools in eighteen months.

The EDI industry has not solved it in sixty years.

Why?

MCP succeeded because it solved the problem for the connector builder, not just the end user.

Clean, simple interface.

Open source spec.

Official SDKs in twelve languages.

Building an MCP server got cheap.

Supply of connectors exploded.

EDI failed to standardise because the complexity was economically valuable to the providers managing it.

SPS Commerce's managed service revenue depends on each connection being complicated enough to require their help. The incentives ran against simplification.

This is the Plaid pattern. Financial data connectivity in 2012 was deliberately hard. Banks had spent decades making it that way. The friction was artificial. The incumbents couldn't remove it without destroying their revenue model.

So a couple of dudes from the outside came in and fixed it, fighting the whole way.

MCP also arrived at a moment when AI made the hardest part tractable. You no longer need a human EDI specialist to map a new trading partner's 856 format. A model that's seen thousands of EDI documents can infer the mapping with high accuracy.

AWS proved this in late 2024 with 50% reduction in development time.

The question is no longer whether AI can map EDI documents, it can, and it will.

The question is whether anyone builds the platform that makes AI assisted EDI onboarding the default experience, rather than an add on feature bolted onto a legacy vendor's existing stack.

Open Sesame

$34 billion growing to $74 billion, apparently.

Legacy vendors economically incentivised to keep it complex. Modern challengers solving specific slices well. Regulatory tailwinds forcing more businesses into digital trading every year.

This category is not a one-winner market.

Orderful owns North American retail and logistics. Or they're on the way to. a16z backed, moving fast, pre-mapped network compounding. If you're a US brand onboarding to Walmart, they probably solve your problem.

Stedi owns healthcare. Deliberately picked. HIPAA's consistent standards make it tractable. Category leader in the making.

AWS owns the infrastructure primitive. Transformation as a service for anyone building on top.

Upstock owns ANZ grocery. Launched 2023 with broader B2B trade ambitions. The site back then talked about general B2B connectivity, multi-partner networks, the whole platform play.

Then reality hit.

They found the wedge that actually paid: FMCG suppliers who just needed Woolworths and Foodstuffs to stop being a six-month nightmare. $40 per connection.

Clean.

Focused.

Now they've got integrations into every ERP built out. The direction is clear. They're going deep on ANZ grocery suppliers.

But that only solves for part of the picture.

Wholesale distributors with 20+ trading partners across multiple regions and protocols. The hospitality group using cXML. The government buyer requiring Peppol. The US retailer mandating X12. The EU procurement system running UBL. The mid-market distributor doing $50M across thirty different trading partners, each with different requirements.

That's the gap adjacent to what Upstock serves.

Orderful owns North American retail.
Upstock owns ANZ grocery.
The complex multi-protocol wholesale distribution layer remains wide open.

The mid-market and SME segment is open.

Suppliers doing $5M to $100M with 20 to 40 trading partners have been structurally ignored by legacy vendors for thirty years because managed service economics don't work below enterprise scale.

This is the biggest underserved slice of the category, and I've pointed out how these teams are itching to get with the cool kids.

Vertical plays beyond healthcare are open.

Foodservice. Hospitality. Construction materials. Wholesale distribution. Each one has enough specific integration patterns that a vertical specialist could carve out defensible territory.

Services businesses wrapped around existing platforms are open.

Most mid-market businesses can afford Orderful or Cleo.
They can't afford the implementation team required to deploy it. Fixed-price onboarding services with domain expertise is a real business that prints cash from day one.

Multiple slices. Multiple moats. Multiple winners.

A Wide Moat

This isn't a simple build.

If it were, the slices would already be filled.

The hard parts aren't the technology.

Technology is buildable.

AI closed the last real gap on document mapping.

The hard parts are the network, the compliance certifications (SOC 2, ISO 27001, HIPAA, Peppol Access Point), the trading partner relationships that take years to build, and the enterprise sales motion that nobody wants to do but somebody has to.

Capital required to build something real in this category runs $50M to $100M over four to five years for a platform play.

Less for a vertical specialist. Much less for a services business.

Orderful raised $43.9M over five years for a mid-scale logistics EDI network. Stedi raised $142M and pivoted to healthcare. A regional or vertical play raises in that range.

The network is the moat.

So What.

If you've lived this, you already know the pain is real. If you say just use API’s, you have not worked as a operator.

Ask any operations manager at any mid market distributor how long their last EDI onboarding took.

Listen for the sigh before the answer.

That sigh is a $34 billion market telling you the current infrastructure isn't working well enough, for enough customers, in enough regions.

Four hotel chains, four protocols, four months each is still happening right now, in real distributors, in real ops teams, in real businesses across many countries.

Still being held together by consultants, spreadsheets, and the institutional patience of operations managers who've absorbed sixty years of unfinished standardisation work as their personal problem.

This is the plumbing underneath global commerce. It's older than cloud computing, older than the internet, older than most of the people building on top of it.

Somebody's going to.