— May 16, 2026 · 10 min read

Open networks vs bilateral EDI: economic analysis

The "PEPPOL vs bilateral AS2" debate is too often reduced to a technical argument. The real question is economic: at what partner volume does an open network infrastructure become structurally cheaper than an N×M matrix of bilateral connections? A pragmatic modelling exercise.

The bilateral model: costs that grow quadratically

In the traditional EDI model, each pair of partners establishes a direct connection: certificate exchange via AS2, parameter negotiation (MIC algorithm, signature, encryption, MDN requirements), integration testing, production. If an enterprise exchanges with N partners, it maintains N distinct connections. For a complete ecosystem of N enterprises where each wants to reach every other, the total number of connections is O(N²/2) — a quadratic growth.

Operational cost of a bilateral AS2 connection typically decomposes into:

• Onboarding cost: initial certificate exchange, gateway configuration on both sides, bilateral tests, MDN validation, first pilot flow. Count 0.5 to 3 person-days per side for a standard partner, more for an OEM or tier-1 retailer with proprietary specifications.
• Annual maintenance cost: certificate renewal (annual or multi-year), incident handling, partner-spec updates, operational support. Count 0.5 to 2 person-days per partner per year.
• Attrition cost: partner or spec change, requiring renegotiation and retest. Hard to model but often underestimated.

The open model: costs amortised across the network

The PEPPOL 4-corner model (and more broadly AS4-based eDelivery networks) shifts the cost structure. Each actor registers once with an access point. Onboarding a new partner becomes negligible: it suffices that both parties are registered in the network's SMP. Certificates are issued by a PEPPOL authority, transport parameters are standardised (AS4 BIS), and discovery is automatic.

Residual costs in a network model:

• Network membership cost: annual subscription to a commercial access point or setup of an internal access point. Count a few thousand euros per year for a commercial AP, several tens of thousands for a certified internal AP.
• Cost per transaction: most commercial APs price on volume, with a degressive unit price. On medium volumes, the order of magnitude is a few cents to a few tens of cents per transaction.
• Standard evolution cost: monitoring PEPPOL BIS evolution and adapting internal mappings. Comparable to the cost of an EDIFACT bilateral version, but mutualised across all network partners.

Where bilateral stays relevant

Three contexts where bilateral retains a clear advantage:

• Very high volumes with a single partner (tier-1 retailers, automotive OEMs, logistics hubs). The marginal AP cost per transaction exceeds that of a dedicated infrastructure, and end-to-end SLA control justifies the investment.
• Deep customisation requirements: partner-proprietary standards (some custom automotive EDIFACT), specific SLAs, sub-second real-time processing. PEPPOL brings standardisation but reduces customisation latitude.
• Strong confidentiality or sovereignty: some sectors (defence, critical energy) refuse to transit through a public or semi-public network. A bilateral AS2 connection over a private VPN remains the reference answer.

The hybrid model dominates in practice

The majority of medium and large enterprises in practice operate a hybrid model: PEPPOL for public partners and intra-EU European invoices, bilateral AS2 for logistics hubs, OEMs and historical retailers, sometimes residual OFTP2 for a few European manufacturers. This duality is not an architectural failure: it is a rational response to different economic contexts.

The modern gateway's role is then to absorb this duality: a single platform that presents a uniform API on the ERP/IT side and that knows how to route to PEPPOL or bilateral AS2 depending on the partner. That is exactly the promise of Stedi, Boomi, MuleSoft or Cleo platforms (covered in the vendors comparative analysis).

Conclusion: think mix, not opposition

The "network vs bilateral" opposition is poorly framed. The two models coexist and will continue to coexist, because they answer different economic contexts. The relevant 2026 choice is not binary but analytical: for each partner family, which model minimises TCO over 5 years taking into account volumes, regulatory constraints, SLA requirements and attrition cost. The integration platform must absorb this diversity.

To dig further, the PEPPOL vs bilateral analysis details the technical mechanics, and the article OpenPEPPOL 2026 gives the country-by-country adoption state.