Skip to content
Interface language: English
Imperium / Signal

A company website outage: a response plan from report to confirmed repair

Prepare roles, evidence and ways to restore the site. The response plan covers availability, forms, the last change and checks after repair.

Article language: English

Two people are beside an intricate mechanism with tubes and pale rounded elements.
In this article

The site opens, the offer looks good, and enquiries never arrive. An outage can keep a perfectly elegant facade. Start by establishing the effect: who cannot complete which task, and since when? Then appoint a coordinator, choose how to restore service and check it from the customer's side. Without that, a repair can easily end with the reassuring message that everything works for the contractor.

The plan need not be long. It needs people, contacts, a few scenarios and evidence that the repair is complete. It must also be accessible when the site or a company tool is unavailable. Keeping the instructions only in the system whose failure prevents them from being read creates a problem that can be removed before the first report.

Identify the scope without guessing the cause

Collect the address, time, description of the user's action and the error message. Ask whether the problem occurs on one device, on a particular network or in several independent attempts. Do not require the customer to provide a technical diagnosis. What they tried to do and what they saw is enough. The person receiving the report should be able to pass this on without losing the address or time.

Separate whole-site unavailability from a function error. A form, a downloadable file and a booking calendar can use different services. A working home screen does not rule out an outage in one of them. Conversely, one local problem does not prove the site is down for everyone. Record the confirmed scope and what has not yet been checked. That distinction helps select the next test.

If a data breach is suspected, follow the appropriate security procedure and involve the responsible people. Do not reduce it to a cosmetic site correction. The first-information gathering principles can be compared with responding to a data incident in an AI tool (in Polish), adapted to the actual event. This article focuses on restoring functions and does not determine incident-related obligations.

Appoint a coordinator and a communication method

One person should coordinate findings even when several teams perform the repair. Their job is to maintain an up-to-date picture: confirmed effect, action underway, owner and next checkpoint. Separately identify the technical specialist and the business decision-maker, for example the person who can pause a campaign leading to a broken form.

Choose where to record actions. An entry should contain the time, observation and step taken. The aim is not a record of every thought. A change trail is needed so two people do not publish conflicting fixes or undo each other's settings. If a conversation happens by phone, transfer the key decision to the shared log before the next operation.

A message to affected people should distinguish fact from prediction. "We have confirmed that form messages are not being delivered; we are checking the sending service" is useful. "It will work in a moment" without a basis creates another expectation. Choose an update frequency appropriate to the effect and communicate changes in state even while the cause is under investigation. The time promised for the next update should be realistic. If the outage delays a particular case, also apply the principles for informing a customer about a delay without a known new date (in Polish). The customer needs to know what to expect before the next update.

Prepare three recovery paths

The first path is to roll back the last change if there is a connection to the outage and a safe return to a tested version is possible. The second is to repair a particular service or configuration. The third is temporary handling of the key process through another route. The choice depends on evidence and effects. A change made yesterday does not by itself prove a cause, but it is an important point to check. When finding causes from evidence (in Polish), separate that timing coincidence from establishing what actually changed the outcome.

For a rollback, record exactly what is restored: code, content, configuration or data. Not every platform covers all of these with one button. If the earlier version uses a service that has since changed, rolling back code alone may not help. The access and materials map from project acceptance (in Polish) should show where the needed elements are and who can restore them.

NIST describes contingency planning as procedures and measures that enable recovery after disruption. It also includes temporarily carrying out a process by another route, including manual means. For a company website, this means preparing an alternative way to handle contact if the primary channel stops working. NIST: contingency planning.

Before using the alternative, check:

  • Does the customer receive a clear explanation of how to submit their case?
  • Is the stated inbox or number actually monitored?
  • Can the team distinguish new enquiries from earlier attempts?
  • Does the scope of data collected fit the alternative channel?
  • After repair, is it clear how to move cases into the normal process?

Preserve evidence before further changes

Keep information about the last deployment, error messages and relevant settings to the extent needed for diagnosis. Do not publicly copy secrets, form data or full dumps containing customer information. The coordinator should know where the material is stored and who can see it. Collecting evidence does not mean distributing entire logs without limits.

Reproduce the error with as small a test as practical. If the problem involves a form, use a controlled message and record its identifier. Check the path from the browser to the receiving location using the B2B form guide (in Polish). One successful technical request does not yet prove that the support team saw the message and can reply.

Aeneia conceptual photography

Make changes with a clear hypothesis and expected result. If you change hosting, the form and domain settings at once, it is hard to tell what helped or made things worse. Time matters during an urgent outage, but an ordered sequence still makes returning to a tested state easier. Each step should have an operator and a condition for stopping further operations.

Hypothetical scenario: enquiries disappear after a change

Imagine a service company receives a report of no reply after a site update. The form displays a confirmation, but the inbox remains empty. This is a fictional example without a real customer's data. The coordinator checks a controlled test and confirms that the problem concerns the sending path, not the entire site.

The team preserves a record of the last change, activates the agreed alternative contact and checks the messaging-service configuration. It does not announce a repair merely because an error disappeared from a dashboard. It sends another test from the public site address and confirms receipt and the ability to reply. It then assesses which earlier enquiries need investigation. Restoring future sending does not automatically explain what happened to earlier attempts.

After stabilisation, the company changes its post-deployment check. It adds inbox confirmation to site acceptance (in Polish) and names an owner for the test in its maintenance plan. Instead of a general instruction to "be careful with updates", there is a concrete action that can detect a similar problem earlier. This is a useful result because it has an owner and a way to perform it.

Confirm the repair and rehearse the plan

The closure condition should match the effect at the start of the report. If a user could not download a file, check the actual address and content. If the site returned errors, check key addresses and server response codes using technical SEO principles (in Polish). If the problem concerned delays, compare the measurement with the earlier state using the performance diagnosis method (in Polish).

Do not finish with "it works for me". Confirm an outcome appropriate to the outage's scope and record the limits of the check. Also decide whether post-repair observation remains and who will perform it. If the fix was temporary, it needs a follow-up task. Otherwise a provisional setting may become the only documentation for the next incident.

Test the plan in an exercise without disrupting production. Choose a scenario, recover the necessary contacts and walk through the rollback decision. Check access to documentation and the deputy's role. NIST's guide emphasises testing to find gaps in recovery procedures. NIST SP 800-34, planning guide. The exercise should expose a gap while it can be fixed calmly.

A contingency plan works more smoothly when you clearly identify the process owner and decision scope (in Polish). Who may stop a change, activate the alternative and confirm a return to work? A list of phone numbers does not itself settle those permissions.

How to prepare a simple first-response card

Put the essential information on one page accessible to people receiving a report. The card should include the website address, key functions, contacts for the coordinator and deputy, and the event log location. Add instructions for collecting evidence without copying sensitive material. A marketing colleague need not know server configuration to describe an undelivered message or an error on a particular page correctly.

For each scenario, identify the first safe test. For unavailability it might be a check from another network and a record of the message. For wrong content, compare the public address with the approved version. For a form, send a labelled test with controlled data. The test should provide information, not change configuration by itself through a person without appropriate access and knowledge.

Also record escalation conditions, meaning when to hand a case to someone with the appropriate decision authority. These might include a confirmed interruption of a key process, a suspected breach or inability to restore service through the normal procedure. Do not set the same alarm for a typo and a loss of customer contact. An agreed scale helps direct attention where the effect is greatest.

Prepare a way to work when the coordinator cannot log in to the main tool. The deputy should know where to find contacts and the latest decision. This does not require copying all passwords into a paper binder. It does require checking that an emergency access procedure exists, has an owner and has been tested within the permitted scope. "Someone has access" is not specific enough.

After the exercise, remove unnecessary steps and clarify instructions. If a participant does not understand an abbreviation, replace it with an action description. If a contact is outdated, correct it and name the review owner. The card should help someone working under pressure. A many-page platform description without a first decision may be technical documentation, but it will not fulfil this purpose.

Review the card after changing the contractor, hosting or form-handling method. Such a change can invalidate contacts and rollback instructions even if the site looks the same. Add a plan update to acceptance of those projects. Readiness to respond then evolves with the site rather than relying on the memory of the person who once published it.

If you need a website with a clear approach to maintenance and acceptance of repairs, see Aeneia Design.

Aeneia / Aeneia editorial

Editorial article. Source links are included in the text.

About Aeneia

Brands we have worked with

  • L’Oréal
  • Garnier
  • Netflix
  • Carrefour
  • Warner Bros. Discovery
  • Biedronka
  • HBO Max
  • Wolt
  • Żabka
  • COSRX
  • Embassy of China in Poland
  • Mixa
Aeneia / Initium

From knowledge to a decision

Tell us what you want to put in order. We will agree on where to start.

Contact Aeneia

We partner with and use solutions from:

  • Google
  • OpenAI
  • Microsoft
  • Stripe
  • Anthropic
  • Linear