Data Migration in Princeton, NJ

Because the move is not declared finished by arrival but by reconciliation, in three currencies: counts (every record class tallied against the frozen source), values (fields aggregated and compared so silent truncation or corruption shows), and sampled human reading (records read side by side in both homes by someone who knows their meaning, the sample chosen adversarially — oldest, strangest, most annotated). Discrepancies are hunted to zero or explained in writing before the old system retires, and the whole reconciliation ships as a document the institution keeps. That is the difference between ‘the import completed’ and a custody transfer you can show a funder: evidence, in the estate’s own file.

It is the normal starting condition, and the method is built for it: discovery catalogues the mess honestly (duplicates quantified, drift located, formats identified, the mystery fields assigned owners to interrogate), and the mapping phase converts understanding into writing — each field’s meaning established with the people who use it before its destination is chosen. The genuinely unknowable fields are not guessed at: they travel with their raw values preserved and their uncertainty documented, so the new home inherits the question honestly rather than a confident falsification. Mess raises the discovery phase’s weight; it does not disqualify the move. What disqualifies a move is pretending the mess isn’t there.

They get a decision, in writing, before anything moves: the archive decision classifies history into what travels live, what rests in a searchable archive (indexed, findable, out of the live system’s way), and what retention policy permits releasing. For most institutions here the resting class is the revelation — decades that must remain answerable but are queried yearly, finally housed as an archive instead of ballast. Provenance travels with every class: the record of what a thing is and where it came from is treated as part of the record, in the archive exactly as in the live system. Losing the archive is the failure the entire method exists to prevent; the decision document is its insurance, signed before the first byte moves.

The calendar is built backwards from yours: discovery and mapping in ordinary weeks, the rehearsal run whenever the data can be spared, the live cutover placed in the quiet window the institution chooses — overnight, between terms, over the dead week — with the old system standing as court of appeal until validation signs off. Day-to-day work continues against the old system during preparation, which is most of the programme’s duration. The honest variable is the estate’s awkwardness (the vanished vendor’s format costs more than the standard platform), and the discovery phase prices it before commitments are made. Disruption, in this method, is a scheduled weekend the institution picks — not a season.

Usually yes — orphaned formats are a specialism rather than an exception: the export’s structure reverse-engineered from the data itself, schemas inferred and verified against the records’ behaviour, the format converted into open ground the future reads. It costs more discovery than a documented platform, and the programme says so plainly at assessment rather than discovering it mid-move. The cases that genuinely cannot move intact are rarer than feared and are identified early, with the fallback designed openly (the frozen system preserved read-only as the archive of last resort, its contents indexed from outside). Either way, the answer comes from examination, not from a sales conversation.

Named engineers, under the engagement’s confidentiality, inside controlled environments — with every copy made during the move accounted for in the transfer’s own audit trail, and working copies destroyed or returned at sign-off as the disposition document specifies. The custody metaphor is operational, not decorative: the same institution that would not hand its archive to an unlogged courier should not hand its database to an unlogged process, and this one is logged. Access narrows further where the material demands (embargoed, personal, privileged classes can be migrated under separated handling), and the arrangement is written down before discovery begins, for leadership’s approval.

Discovery and the mapping document are a defined engagement — and they are where honest pricing becomes possible, because migration cost lives in the estate’s specifics: volume, formats, duplication, the mystery fields’ count. The execution is then priced against the written mapping, rehearsal included, so the number leadership approves describes the move they are actually getting. The comparison is the improvised version’s true price: the same move done casually, plus the discrepancies discovered later by the wrong audiences, plus the re-move when the first one’s proof cannot be produced. Custody done once, with evidence, is the cheap version — it merely looks thorough on the invoice.