With GrantHub’s sunset date passed on 31 January 2026 and the accounts behind it closed, most organizations are left with a folder of downloaded files as their entire record: years of submissions, funder contacts, narrative libraries, and deadline calendars. Whether that folder is complete is a question usually answered too late. A migration truly begins not when you shortlist a replacement but when you audit what came out of the old system.

Stage one is inventory. Write down in plain language what you believe lived in the system, with rough numbers beside each category. You will be wrong, and the gap between your guess and reality is the first thing to investigate. This is also the moment to check your retention obligations under the Uniform Guidance for federal awards, which generally run from the submission of the final expenditure report. Where it goes wrong: nobody inventories anything, so the export silently defines what existed, and anything it omitted becomes invisible.

Stage two: gather every copy that was taken, store the raw files untouched, and compare snapshots to see which fields were volatile and whether anything truncated on the way out. Treating one late file as authoritative means a flaw that a fresh request once would have fixed is now permanent.

Stage three reconciles the export against outside evidence. Sample awkward records across your history, not just recent tidy ones, checking award letters, funder portals, and the general ledger. Old records carry the conventions of whoever set up the system, and those are the rows that break an import.

Stage four rebuilds relationships, not rows. Grant records form a graph of funders, opportunities, applications, awards, payments, and reports. Flat exports flatten it, and matching on organization name creates duplicates. Any replacement should show its import mapping field by field before you sign. Where it goes wrong: row counts match while payments attach to the wrong award year, unnoticed until a reconciliation report runs.

Stage five recovers attachments with context. Bulk retrieval is the least reliable part of extraction and cannot be redone. A folder of thousands of PDFs with no dependable link between file and submission costs more staff time to rebuild by hand than the rest of the migration combined.

Stage six reconstructs obligations, not just dates. A deadline is a date; an obligation adds a requirement, an owner, a format, and a consequence. Federal work compounds this, since Grants.gov workflows and SAM.gov registration renewals continue regardless of what happened to your tracker.

Grantcue (https://www.grantcue.com/blog/grant-proposal-statement-of-need-examples-template)

Stage seven decides what happens to people and history. User attribution is the closest thing a small shop has to an audit trail, and deactivating departed staff may strip their names from records. Decide deliberately whether to carry it, reconstruct it, or accept its loss, and write the decision down where a future colleague will find it.