You can sign in again after an outage. The supplier says the service is restored. But an employee asks, “What happened to the invoice I sent just before it stopped?” That question marks the difference between a working screen and finished business recovery.

In this example, the invoice might have been accepted, might still be waiting or might not have reached the service at all. Before sending it again, your team needs a way to find out. Clarifying that procedure before an outage is much easier than working it out under pressure.

An outage tells you that some work is unavailable. It does not tell you why. A failed component, an accidental change or a deliberate attack could each interrupt a service; the interruption alone does not choose between them. A cybersecurity incident is a security problem that requires response, not simply another word for every slow page. NIST (the US National Institute of Standards and Technology)‘s framework distinguishes potentially adverse events, investigation and declaring an incident when the defined criteria are met. Let the supplier explain what has been confirmed rather than guessing the cause from a failed login. 2

There are also different harms to ask about. Can we use the service when needed? This is availability. Are the records accurate and protected against improper changes? This is integrity. Have the records been exposed to people who should not see them? This is confidentiality. A service could be unavailable while stored invoices remain intact and private. It could also let you sign in while some records have been changed or exposed. These are separate questions, even when one incident raises all of them. NIST’s incident-response guidance treats these security concerns separately rather than reducing recovery to reopening the screen. 5

First find out which part stopped

The portal your staff use, the connection from your accounting program and the onward document service are different parts of the journey. A planning exercise might consider a portal outage while onward processing continues, or a failed accounting connection while the portal remains available. A message saying “the service is down” does not explain which work is affected.

Ask the supplier to explain the next action in terms your accounts team can use: keep working, pause submissions, use an agreed alternative or wait for confirmation. Also agree where updates will appear if the usual portal is unavailable. A support route that exists only inside that portal is not enough for this scenario.

An initial update can be useful without knowing the final cause. “We are investigating, this function is affected, and the next update will follow at the agreed time” gives customers something to work with. These are suggested communication arrangements, not universal legal reporting deadlines. NIST’s incident-response guidance treats preparation and learning as parts of response, not activities to postpone until everything has failed. 5

A backup is the beginning of recovery, not the end

A backup is a separately kept copy of information intended to help restore it if the working copy is lost or damaged. It is like a spare copy of a book. Having the copy matters, but you still need to know whether the pages are intact and whether it is the edition you need. For a business service, someone must also restore the systems that make the records usable.

NIST’s Cybersecurity Framework includes verifying backup material before using it for restoration and checking restored assets before confirming normal operation. In ordinary language: do not equate “we have a copy” with “you can reliably resume work.” 2

Consider the original uncertain invoice as a recovery example. The accounting program sent it, but the confirmation did not return before the interruption. The supplier later restores an earlier copy of its records. Your source invoice can still exist in the accounting program while the supplier’s newest processing record needs reconstruction. Restoring stored information and finding out what was sent during the intervening period are therefore connected but different jobs. This example describes a possibility to plan for, not something measured at a provider.

An older copy can be undamaged yet incomplete for today’s work. A newer copy can contain a problem that had already occurred before it was made. Checking integrity is not merely checking that there are files in a folder; the recovery team needs reasons to trust and use the restored material. NIST explicitly includes checking backups before restoration and restored assets afterwards. Ask the supplier to explain how this leads to a usable service for the functions you bought. 2

There are also different promises hidden in “we can recover quickly.” One concerns how long work may be interrupted. Another concerns how much recent information might need to be recovered or reconstructed. Ask for both in plain language, and distinguish the supplier’s target from the commitment in your agreement and the result of any actual exercise.

Keep this tied to what you bought. If the supplier retains documents for you, recovery should address those records and their usable context. If it only transports documents, the discussion concerns that transport function and unfinished work. Not every invoicing service is a document archive.

Resolve the uncertain invoice before restarting blindly

Return to the employee’s question. A practical recovery explanation should identify where to find the invoice’s state and what that state means. “Accepted for processing,” “sent onward,” “rejected” and “still waiting” describe different stages; the supplier must explain the actual labels and records in its product rather than relying on these example words.

Imagine two records for our example: the accounting program says “submitted,” and the supplier’s processing history says “accepted for onward handling.” Neither label alone explains whether the destination received the document. The supplier should identify the available confirmation and its meaning, or say that the outcome remains unresolved. Keep the original reference—the service’s label for this submission—so support can discuss the same operation. A later repeat can otherwise become a second piece of work rather than clarification of the first.

Your team needs a business conclusion for that invoice: confirmed result, permitted repeat under the agreed procedure, or a held item that someone is still investigating. This conclusion is what turns technical restoration into a manageable return to work. It also prevents an unresolved item from disappearing merely because the homepage is available again. The example words here are not standardized status promises; the actual service must explain its own records.

Agree who compares the accounting program’s submission record with the supplier’s processing record and resolves any disagreement. That comparison is called reconciliation. The technical team may restore the system, while your accounts team still needs to confirm what remains to be done. Neither should assume the other has completed that check.

You can rehearse this safely in a supplier-authorized test environment. Use a made-up invoice and temporary accounts to simulate an uncertain submission, then have the supplier explain whether and how it should be repeated. Agree small volumes and stop conditions. Do not interrupt production, use live customer data or create real chargeable messages. The purpose is to understand the documented procedure, not to manufacture an actual outage.

Leaving requires some of the same preparation

Recovery asks, “What do we need in order to continue?” Changing provider asks a closely related question. You may need invoice files, attachments, processing states and explanations of what the fields mean, depending on the service you purchased. A download button does not show that the next system can use the result.

Ask for an agreed sample export containing only made-up records. Have the person who would receive it explain whether it is readable and useful. If your accounting program already holds the authoritative invoice record, the supplier’s processing history may be the missing part. If you rely on the supplier for retention, your handover needs will be different.

NIST’s framework explicitly includes planning for activities after a supplier agreement ends. Use that idea while the relationship is cooperative: discuss transition assistance, its price, access during the change and how old access and retained data will be handled. These are arrangements to clarify, not legal rights this article assumes you already have. 2

Peppol also distinguishes the services that carry documents from components used to find where they should go. If your change involves Peppol, ask the old and new providers to explain the relevant handover steps. Cancelling a portal subscription is not a complete migration plan. 6

After a disruption or rehearsal, keep the lessons that change what people do next time. NIST encourages using incident lessons to improve practice. For your team, the most useful conclusion may be simple: who checks unfinished invoices, where the instructions are and whom to contact when the usual screen is unavailable. 5

Sources