1. Describe the website you need to recover
Start with a concrete inventory: pages and translations, media, downloadable documents, data managed in the content tool, code and settings needed for operation. For each item, record the copy’s location, date, owner and recovery procedure. If an item does not exist in your project, mark it as not applicable.
Ask whether images and content are kept with the code or elsewhere. A code repository may contain neither files uploaded by the team nor site data. Record external dependencies separately: domain name, hosting and connected services. This proposed inventory helps you ask questions; it does not describe what your hosting provider supports.
2. Set two objectives before choosing a frequency
NCSC NZ distinguishes the RPO, a target for acceptable data loss measured in time, from the RTO, a target for recovery time. These objectives guide backup decisions; they are not guaranteed outcomes.
In a fictitious example, a team accepts re-entering one day of page changes and wants its website back within four hours. Write down both objectives, then ask what the proposed solution actually covers. If customer enquiries are stored elsewhere, address their recovery separately. Daily backups do not prove the latest copy succeeded or that four hours will be enough to restore it.
3. Plan for a protected copy and usable access
The Canadian Centre for Cyber Security recommends encrypted backups, a secure external location and access limited to people responsible for testing and restoration. NCSC UK stresses knowing how to restore and checking that important data is present.
For your preparation, ask how to reach the copy when the normal website is unavailable. Name an owner and an authorised alternate. Document how to recover decryption material without copying passwords into the test sheet. Clarify retained history and exclusions. An exercise should uncover a missing dependency before an emergency.
Source : Canadian Centre for Cyber Security — Back up and encrypt data · Source : NCSC UK — Backing up your data
4. Prepare an isolated restoration environment
Agree with the maintainer on a space separate from production, with restricted access and the necessary resources. Before importing anything, confirm the destination and the absence of connections that could overwrite the real site. If personal data is needed for the exercise, limit its use and access; prefer fictitious data when it is sufficient.
Disable or replace outgoing email, payments, notifications and synchronisation in this environment. A test page must not trigger a real action. Do not rely solely on the absence of public links or search-engine instructions to protect the copy. If isolation cannot be established, stop the exercise and have a suitable environment prepared.
5. Restore an identified copy and check the journey
Select a specific backup and keep its date alongside the expected site version. Have an authorised person follow the documented procedure. Measure the time until the restored site has been checked, including manual steps. Record blockers, missing access and unavailable files. Keep a partial restoration labelled as partial.
In the isolated environment, open a main page, a recently changed page, an image and a download. For a multilingual site, compare each language with the backed-up version, then test navigation on a phone. For an interaction, use only a test destination whose isolation has been confirmed. Do not use a real customer enquiry as evidence that it works.
6. Keep a report and decide what comes next
Compare recovered content and measured time with the written objectives. Use the checklist below to retain understandable evidence: selected copy, expected result, observed result, duration and defects still to fix. Opening one page does not by itself validate the expected media, documents or data.
Assign each correction and agree with the maintainer on a regular exercise schedule, as well as a fresh check after a significant website or hosting change. Keep a report without secrets, then have the test copy removed or protected under your team’s rules. Any later production recovery needs a separate plan that accounts for changes since the backup. After a suspected attack, first have the incident and the copy’s trustworthiness assessed before putting the site back online.
An evidence sheet for your exercise
| Check | Expected result to agree | Evidence to retain |
|---|---|---|
| Scope | Necessary content, media and settings identified | Inventory and exclusions |
| Selected copy | Known date and expected version | Reference without secrets |
| Isolation | No writes or messages to real services | Environment validation |
| Pages and files | Sample consistent with the selected copy | Addresses and observations |
| Measured recovery | Loss and time compared with objectives | Timeline and gaps |
| Next steps | Owners and next check defined | Dated report |
Frequently asked questions
Is a hosting provider’s backup enough?
Ask what it contains, how many versions are kept, who can retrieve them and how restoration is tested. Decide from those answers and an exercise, without assuming every plan covers the same scope.
Should I test directly on the public website?
For the exercise described here, use an isolated environment and test destinations. Actual production recovery is a separate operation to prepare with its owner and an appropriate plan.
Can one test establish a recovery-time promise?
A test provides an observed time under specific conditions. Keep those conditions and limitations; do not turn the result into a guarantee for every incident.
Prepare your first restoration sheet
Choose an important journey and ask the maintainer to complete the inventory, objectives and isolation conditions. Then have the copy to test approved before scheduling the exercise.
Sources and scope
Sources consulted on 4 November 2026. Attributed recommendations are linked in the relevant sections. The inventory, numerical example and checklist are proposed working methods, without an audit of your installation or a service commitment.
