Content reviewed: 12 October 2026

Retention, reducing active storage and restoring usable data are different objectives. Identify the required outcome before choosing a feature, product and licence.

Retention controls the content lifecycle

Retention policies and labels can preserve or delete content under configured conditions. Location, scope and the start of the period matter. Microsoft explains the model in its retention overview.

“Keep everything for seven years” does not define the starting point or the final action. A period starting at creation differs from one starting at the last modification. Supported settings vary by workload.

Use real content examples when designing the policy. A contract amended over several years has a different lifecycle from a one-off staff notice.

Seven years from which starting point

Illustrative document: created on 15 March 2020, modified on 15 March 2026. A seven-year period from creation points to 2027; from modification, to 2033. If the required start is contract expiry, creation date is insufficient: the retention model also supports event-based labels with their own prerequisites. This is an arithmetic example, not a recommended legal period or a promise of deletion on that exact day.

SharePoint and OneDrive retained copies can occupy the Preservation Hold Library and consume site quota. Retention may preserve content after removal from the ordinary view without releasing storage. Location policies and item labels also behave differently when content moves. Microsoft explains these distinctions; include the new file location in testing.

An archive is different from a backup

An archive supports older content or moves it outside active storage. It does not establish how quickly or completely the business can recover from a destructive change.

For email, “archive” may mean an online mailbox archive. For SharePoint, it may mean Microsoft 365 Archive. These are different products and workflows; use the exact name in the design.

Specify who needs access and whether they need individual documents or regular work across all the content. That operational requirement affects the choice.

What Microsoft 365 Archive does

Microsoft 365 Archive moves inactive SharePoint files or sites into a cold tier; it is not a historical recovery point. Content remains searchable, but normal direct access requires reactivation. Do not assume a Teams team archives as one unit: private and shared channels have separate sites and limitations. Check the site template and client support before changing tiers.

Evaluate backup through recovery

Microsoft 365 Backup protects supported Exchange Online, OneDrive and SharePoint data. The current service overview describes restore options and limitations.

Ask what needs restoring, to which point in time and within what deadline. A “backup completed” status does not answer those questions.

Restoring an entire site, selected messages and an archived document are different tasks. Check supported scope, administrator permissions and effects on newer valid data.

Compare the purpose of each layer

Mechanisms and recovery checks
LayerPurposeTest
RetentionPreserve/delete under rulesScope, period start and discovery
Version historyPrevious file statesAvailable versions and trimming
Recycle binRecover deleted contentObject availability within the supported window
ArchiveStore inactive contentAccess/reactivation and cost
BackupRestore protected dataRestore point, granularity and elapsed time
Service resiliencePlatform availabilityIncident procedure and business dependencies

These mechanisms can complement each other. Test their outcomes rather than treating them as interchangeable.

Backup scope and available points

The current overview defines a SharePoint site, OneDrive account or Exchange account as the core protection unit. An in-place site or OneDrive restore rolls content and metadata back and overwrites later changes; a new URL is another option. Exchange recovers modified or deleted items. “Restores all of Microsoft 365” therefore does not prove coverage of every related application or setting.

The overview now lists policy recovery windows of three or six months, one year or two years. The FAQ still describes some scenarios as one year, while the overview’s point table ends at 52 weeks. Check the actual policy and restore calendar; a two-year window does not establish ten-minute points throughout both years. The FAQ distinguishes ten-minute SharePoint/OneDrive points in the trailing two weeks from weekly older points.

Resilience does not prohibit every deletion

Microsoft 365 Backup uses append-only mechanisms to prevent overwriting stored points, but explicitly allows backup deletion during offboarding. It describes a recovery grace period for existing backups and multi-admin notifications. Include Backup administration permissions and accidental protection removal in the design; do not interpret append-only as an absolute prohibition on deletion.

Start with an incident scenario

Choose a specific incident: deleted folders, mass file overwrites, an integration error or the need to produce historical correspondence.

List affected data, acceptable data loss and the recovery deadline. Then check which available mechanisms address the scenario.

Management can approve a concrete outcome. “We back up Microsoft 365” is too broad if required workloads or objects are missing.

Checking work data on laptop

Example: an overwritten document library

A hypothetical company stores sales proposals in a library. Automation overwrites many files, and the problem is discovered several days later. IT must recover valid content while preserving subsequent legitimate work.

Before restoring, identify the last healthy state, damaged scope and newer valid changes. The administrator compares versions and supported restore options; the sales team checks the recovered proposals.

The example does not prescribe one universal product. It explains why a restore test must handle the conflict between an older recovery point and newer valid work.

Extend the library scenario: damage occurred at 09:00 Monday, was discovered Wednesday, and ten valid proposals were created meanwhile. Restoring the whole site in place to Monday’s state can remove those later changes. Preserve the valid new work before proceeding and consider a new-URL restore where supported. Acceptance must confirm both repaired files and retained proposals; automatic merging of the states is not guaranteed.

What RPO and RTO mean

RPO describes acceptable data loss, expressed as time. RTO is the target time to resume operations. Explain them as how much recent work may be lost and how long the team can wait.

Do not derive RTO from transfer speed alone. Recovery includes approval, administrator access, point selection, execution and content-owner verification.

Record total elapsed time and each stage. A missing decision-maker can delay recovery even when the technical restore is fast.

Retention licensing depends on the feature

Purview supports several retention and label mechanisms. Manual application, automatic deployment and advanced functions have different requirements.

Use the Purview service description. Identify the exact feature and the covered users or locations; portal access alone is not licensing evidence.

For organisation-wide protection, include employees using different suites. One advanced administrator licence does not automatically cover everyone benefiting from a feature.

Examples of licensing differences
RequirementCheck
Site or mailbox retention policySupported plans for that location; broad retention is different from Teams retention.
Manually published labelPublishing rights and users benefiting from the feature.
Adaptive scope or automatic label deploymentHigher licensing requirements for the method; E3 is not automatically sufficient.
Event-based period or disposition reviewAdvanced records-management rights, not just ordinary label creation.
Microsoft 365 BackupConsumption-based PAYG, separate from per-user Purview licensing.

The Purview licensing description includes SharePoint site owners and members as feature beneficiaries; visitors are treated differently in that stated scenario. Do not derive quantity solely from administrators. Document the specific feature and role instead of a general “retention for everything” requirement.

Run a meaningful recovery test

  1. Create safe representative test content.
  2. Verify it is inside the intended protection.
  3. Simulate a change or deletion.
  4. Restore the required scope through the supported procedure.
  5. Check content, metadata and access.
  6. Resolve newer changes and conflicts.
  7. Have the business owner accept the result.

Record failures as well as successes. Eventually finding one file does not prove that the complete business process can be restored.

Frequently asked questions

Can retention replace backup?

Decide by the required outcome. Retention and backup have different purposes; validate the specific recovery mechanism.

Does Microsoft offer Microsoft 365 backup?

Yes. Microsoft 365 Backup is a separate service; verify its supported scope and settings.

Must backups be outside Microsoft?

That depends on required separation, risks and recovery objectives. The vendor name alone does not establish suitability.

Does seven-year retention preserve everything for seven years from today?

No. The start of the period, scope and other rules determine the result.

Keep an actionable design record

Retain the protected-data map, responsibilities, licensing evidence and tested recovery scenario. These records support a decision better than a list of enabled settings.