Do You Need a Backup EDI VAN?
Learn whether you need a backup EDI VAN, what breaks during an outage, and how to build an EDI disaster recovery plan that survives a real test.
Most companies need a documented failover plan, not necessarily a second paid VAN contract. The exception is if you run just-in-time (JIT) automotive feeds, high-penalty retail compliance programs, or healthcare transaction sets where even a few hours of downtime creates real damage. For everyone else, a tested backup and restore plan plus a secondary connection method covers the risk. EDI is a mission critical component for most organizations, which means a disaster recovery and a failover plan should be in place, and that single sentence from the EDI Academy's networking brief is really the whole argument for why you can't treat this as optional, even if "backup VAN" specifically isn't the right answer for your environment.
What Actually Happens to Transactions During a VAN or EDI Server Outage?
It depends on where the failure sits. A VAN mailbox holds queued documents until your system reconnects, so a short outage on your end rarely means lost data. A failure on a direct AS2 or SFTP connection is a different story, because there's no third-party buffer holding the interchange while you're down.
The GraceBlood team describes a case where a client, a $60M company, asked for emergency assistance after their EDI server crashed and they had no redundancy and worse yet, no backups. Their framing is blunt: recovering from an EDI server failure is time-consuming and unpredictable, and at a minimum at least one of these fail-safes will minimize the risks of elongated downtime. On the backup side specifically, there is no excuse in 2023 for not performing backups whether it's for your PC or your EDI server, and keeping backup files in a separate location means that if your server crashes, your backup data stays safe and you can get back into production quickly.
What should be backed up isn't just the database. With EDI in place it's important that data is backed up frequently, and all EDI data formats should be backed up, meaning the machine-readable EDI data, the translated proprietary data, communication logs and the database itself. Lose the communication logs and you lose your ability to prove to a trading partner that you sent the 997 or 999 functional acknowledgment on time, which becomes an argument you don't want to have during a chargeback dispute.
Is a Direct AS2/SFTP Connection Riskier Than a VAN During an Outage?
Yes, in one specific way: direct connections shift the burden of redundancy entirely onto your infrastructure and your partner's, while a VAN provides a buffering layer by design. That's the tradeoff IT directors are making, often without realizing it, every time they migrate a partner off a VAN onto straight AS2 to cut per-transaction fees.
AS2 introduces its own failure points that a VAN abstracts away: expired certificates, TLS version mismatches after a partner retires an old cipher suite, or a partner-side outage with no queue to catch the documents in between. That's part of why some IT directors deliberately keep one or two of their highest-volume partners on a VAN even after migrating most of the trading base to AS2 or API. The buffering isn't about cost anymore at that point, it's an insurance policy.
Managed providers like SPS Commerce, TrueCommerce, Cleo, IBM Sterling, and EDICOM build redundancy into their infrastructure so you're not maintaining failover hardware yourself. The same logic shows up downstream on the execution side of the supply chain: platforms like MercuryGate, Descartes, Transporeon, Alpega, and Cargoson maintain redundant carrier API connections specifically so a single carrier outage doesn't stall load tendering across your whole transport network.
How Much EDI Downtime Can a JIT Automotive or Retail Supplier Tolerate?
Effectively zero for sequenced automotive JIT, hours to a day for most retail or DC replenishment flows, and longer for lower-frequency transaction sets like 820 remittance advice or 830 planning schedules. The tolerance isn't a single number, it's set by how close the transaction sits to the physical line.
The EDI Academy's framing is direct on this point: the speed at which the EDI environment needs to become operational again will dictate the type of fail over requirements, and the EDI failover time is zero in a sequenced manufacturing scenario where parts arrive on an hourly cadence feeding the assembly line. There's no "we'll catch up tonight" option when a missed 862 JIT call means an idle line in an hour. Retail replenishment and DC-to-store feeds have more slack, which is exactly why the DR conversation needs to happen per transaction set, not once for the whole EDI environment.
What Should an EDI Disaster Recovery Plan Actually Include?
Three distinct things, not one: failover, backup, and disaster recovery, and most plans fail because people treat them as interchangeable. Failover is the automatic or manual switch to a standby when hardware dies. Disaster recovery is the broader response if a plane crashes into the data center building, and a well-implemented plan includes all components of the IT environment and not just EDI, because your EDI server is useless if the ERP it talks to is still offline.
Set your targets using the standard pair of metrics: recovery time objective (RTO), the maximum downtime you can tolerate in a disaster scenario, and recovery point objective (RPO), the maximum data loss you can tolerate, as Microsoft's reliability documentation frames it. A workable plan includes:
- Trading partner specifications and maps stored offsite or in a separate cloud repository, not only on the production EDI server
- A mirrored or standby EDI server environment that can take over without a full reinstall
- Offsite or cloud backups covering raw interchanges, translated data, and communication logs
- A defined RTO for each transaction set, since a missed 862 and a missed 820 are not the same emergency
- Escalation contacts at each high-volume trading partner, agreed before the outage, not during it
During disaster recovery testing it makes sense to confirm that in the off-site data center location, the DR business application server and the DR EDI server successfully interact with each other, because a standby EDI server that can't talk to a standby ERP instance isn't a recovery plan, it's a false sense of security.
Should You Run Two VAN Providers at Once, or Just Keep Backups?
Full dual-VAN redundancy is rarely worth the cost and complexity for most shippers. It's reserved for the highest-penalty, highest-volume compliance programs, the kind already covered on this blog around major grocery and mass retail routing guides. For everyone else, the cheaper middle ground works: a single VAN or AS2 provider, a documented manual fallback such as portal entry or an EDI-to-email bridge for critical POs, and offsite backups of your maps and trading partner agreements.
| Approach | Setup cost | Who owns redundancy | Best fit |
|---|---|---|---|
| Single VAN | Low to moderate | VAN provider | Most mid-market shippers |
| Direct AS2/SFTP, no backup link | Low | Your IT team | Low-penalty, low-volume partners only |
| Managed EDI provider (Cleo, TrueCommerce, SPS Commerce, IBM Sterling) | Moderate | Provider's infrastructure | Companies wanting to offload network risk |
| Dual VAN | High | Shared across two providers | Highest-penalty retail/automotive compliance programs |
How Do You Test an EDI Failover Plan Without Breaking Production?
Use a sandbox trading partner ID, schedule drills during off-peak windows, and agree staged cutover times with key partners in advance, the same discipline used for peak-season load testing. You're not trying to prove the plan works in theory, you're trying to find out where it breaks before a real outage does.
AWS's disaster recovery guidance puts it well even though it's written for cloud infrastructure rather than EDI specifically: disaster recovery testing should focus on how the workload reacts to loss of a region, whether traffic routes away from the failed region successfully, and testing for a data disaster is also required. Translate that to EDI terms and it means testing both your network failover and your actual document integrity after restore, not just whether the server boots.
Start small: pick your single highest-penalty transaction set, run a failover drill on it next quarter, and measure how long it actually took against the RTO you wrote down. If the number embarrasses you, that's the plan working exactly as intended.