Your overseas launch team wants a packaging shot replaced before the next campaign. The production partner in China still holds the footage, so the request appears to be a small editing job. But ownership of a file does not tell you when an editor can actually use it. Retrieval, transfer and project recovery may consume the available window before the first edit begins.
For ONCE, that changes the procurement question: which archive service brings an agreed set of assets back into an editable state by your deadline? Storage capacity alone cannot answer it.
A retrieval estimate is only one part of the schedule
Amazon S3 provides a useful, specific example. Its documentation, checked on October 31, 2026, lists typical Deep Archive retrieval times of within 12 hours for Standard without Batch Operations and within 48 hours for Bulk. Deep Archive has no Expedited option. Large restores can take longer. These are service-level retrieval descriptions, not promises about a finished campaign video.
The names also matter: Glacier Instant Retrieval supports real-time access without a restore operation. A supplier saying “Glacier” or “cloud archive” has not yet specified your access model. See the official retrieval options and archived-object guidance.
These facts describe AWS, not every production vendor or an industry-wide transition. The procurement method below is ONCE analysis, intended to help buyers evaluate whichever storage service their partner actually uses.
Buy an editing-readiness clock
Ask for four milestones, with an owner and observable completion event for each:
- Selection: the project version, required media and access permissions are agreed.
- Retrieval: the listed cloud objects are readable; failures and any availability expiry are reported.
- Transfer: the agreed working destination receives the files with an integrity check.
- Project recovery: the editor opens the project, reconnects essential media and exports an agreed sample.
A conservative planning allowance adds those four stages. Where work overlaps, use the actual critical path instead. Editing, client review and release need additional time. Define when a valid request starts the clock, record client-dependent pauses explicitly, and express deadlines with a time zone. “Restored today” should never mean merely “a restore request was submitted.”
Compare the cost of the same retrieval task
Request separate allowances for retention, retrieval operations and data, temporary working storage, transfer, and human effort to locate, reconnect and verify assets. Mark included services so the same work is not charged twice. Some storage designs will not incur every item.
There is a further handoff risk. AWS documents temporary restored copies for Flexible Retrieval and Deep Archive, with a specified availability period; archive and restored-copy storage can both be billed. Read Restoring an archived object. Your contract should identify who moves the working files before expiry, who extends availability, and who approves additional charges if a campaign slips.
Compare bids using the same asset list, deadline and working-copy period. ONCE is not quoting cloud prices or asserting an average production cost here.
Keep the next likely change accessible
Consider a hypothetical appliance brand sourcing products and production in China. Its next campaign may need a regional plug, revised packaging and a different language track. Keeping the relevant project, approved master and necessary shots accessible may matter more than keeping every historical take equally close at hand.
For rarely reused footage with a long planning window, slower retrieval may be acceptable. A proxy can help select a shot but cannot replace the source needed for finishing. Build the accessible set around a plausible modification, then agree what happens when the request exceeds it. This belongs in the scope of a video production service, alongside responsibility for the edit itself.
Use a bounded restore exercise before renewing
Choose one agreed older project and a small, specified change. Record request time, storage class, retrieval method, data size, temporary-copy expiry, transfer completion and project recovery. Require a checkable sample export plus an exception list. This is a proposed acceptance exercise, not a report of an ONCE client test.
If permissions, missing media or unavailable plugins block the edit, faster retrieval will not fix them. If retrieval is the bottleneck, consider a different access tier, advance retrieval or a revised commitment. A managed archive should earn its renewal through demonstrated readiness and clear responsibilities. Over time, those records can inform what deserves accessible storage and what can wait.
Put the decision into the brief: identify the assets that can return to editing by our campaign deadline, then price and demonstrate that service. That gives the buyer a usable basis for choosing a retention-only arrangement or a more demanding production retainer.
Source note: the three AWS pages above do not state publication dates; checked October 31, 2026. Timing examples are typical values for the specified options. The planning framework and hypothetical brand scenario are ONCE analysis, not measured results or forecasts.
