The challenge
The agency’s entire working history lived in Dropbox: 2.7 terabytes, 3.3 million files, 585,000 folders. For a creative production business that archive is the asset: every campaign, every shoot and every version of every deliverable.
The problem was not storage. After 585,000 folders, finding anything depended on whoever created the folder having named it sensibly, and on the searcher knowing what that person was thinking. Retrieval had quietly become an oral tradition.
Two constraints made the obvious fix awkward. The agency runs on Macs, so any answer had to include working desktop sync rather than a web portal people would resent. And the archive was already tiered in practice: a small body of active work, a larger body of recent but idle material, and a very large cold archive that must stay findable and almost never changes.
The solution
The content took two hops on purpose. Dropbox to SharePoint Online first, then SharePoint Online to an on-premises SharePoint farm. That looks like a detour and it is the entire architecture. Microsoft 365 gives the agency Mac desktop sync, sharing and collaboration; the on-premises search service gives it what Microsoft 365 could not: a full-corpus index across all three storage tiers with custom facets, on hardware sized for the crawl rather than metered by it.
Search was built as two products for two needs. Hot content got faceted search with three custom filters on top of the standard set; warm and cold storage got out-of-the-box search. That split was priced separately in the original estimate, 8 hours for the hot experience and 4 for the rest, because building that interface across a cold archive nobody browses would have been spending the client’s money on a demo.
Metadata was created during the move, not promised after it. A custom migration script populated enterprise metadata columns as documents landed. Migrations that defer tagging produce a new system with the old system’s findability problem.
A proof of concept ran against the real corpus first. The whole tenant was scanned to establish the true numbers, a subset of warm and cold storage was crawled, and roughly 8,000 documents were pushed through the full two-hop path with unified search over them. Every number in the plan afterwards came from measurement.
The estimate said what it would not do. OS installation, domain joins, ongoing infrastructure maintenance, the Sharegate migration licence and Microsoft licence costs were listed as exclusions in the 120-hour quote. The agency’s existing managed provider kept that work, and the boundary never had to be renegotiated.
The results
Finding a three-year-old asset stopped depending on remembering who filed it. Producers search by facet, on client, campaign and asset type, across hot, warm and cold storage in one place, while their Macs sync live work through OneDrive exactly as the Dropbox client used to.