Adobe Experience Cloud
Five million AEM assets need metadata updates without overwrite
Plan an append-safe AEM 6.5 metadata migration with explicit merge rules, pilot batches, and recovery.
October 29, 2026
An AEM 6.5 on-premise team needs to update roughly five million assets and about twenty metadata properties from a mapping file while preserving values already present. Those figures come from an Experience League request, not from an Adovo engagement. At this scale, “append” is not a single operation: single-value fields, multi-value fields, blanks, and conflicting values each require a rule.
Write the merge contract before the import
For every property, specify whether a source value fills only empty targets, replaces a stale target, or adds a distinct item to a multi-value list. Decide how to compare case, whitespace, and duplicate values. Define whether an empty source cell means “leave unchanged” or “clear.” Record asset paths and a version of the mapping file so a rerun is deterministic. A batch job without these semantics can preserve technical uptime while corrupting editorial meaning.
Validate Adobe’s supported bulk route on a sample
Adobe’s AEM metadata import/export guidance describes CSV shape, including the assetPath column and property types. Test a representative sample with existing values, missing values, multi-value properties, and special characters. Export before and after states and compare each field. The documented import path can update metadata in bulk; it does not itself certify that a particular CSV implements the team’s append policy.
Scale in checkpoints, not one giant run
Partition by stable asset path or folder, schedule modest batches, and measure author response, write throughput, failures, and backlog. Log per-asset outcomes and make reruns idempotent. Preserve a rollback export or versioned snapshot appropriate to the repository. Confirm with AEM operators whether workflows or indexing need scheduling changes in this on-premise environment. Promote to millions only after a pilot proves value preservation and recovery, not merely a green completion notification.
Adovo’s view
Bulk metadata work is a data migration wearing a DAM interface. The hard part is defining which existing knowledge must survive. Make preservation rules testable on ten difficult assets before multiplying them by five million.
