Switching feed providers is one of the few genuinely dangerous routine operations in ecommerce. Google's PMax learning is keyed to your offer ids: if the new feed ships the same products under different ids, Google sees a brand-new catalog, the learning history resets, and performance craters for weeks. The migration flow exists so that never happens by accident.
Why identity has to match
An offer id is how Merchant Center knows that the item in the new feed is the item it's been serving for months. Simprosys and other providers each have their own id conventions — plain variant ids, prefixed, suffixed per market. If Magistry's feed doesn't reproduce your current convention exactly, every product becomes a stranger to Google on cutover day. So the first phase of every migration is establishing exactly what the live convention is.
The parity analysis
Before anything changes, Magistry runs an offer-id parity analysis: it compares what its own feed would ship against the offer ids currently live in Merchant Center, item by item, and reports the match rate with every mismatch listed. You read that report before any cutover step is available. A parity gap isn't a warning to click past — it's the list of products that would lose their history.
Gated phases, operator-confirmed
- 1Analysis — the parity report is generated. Nothing has changed in Merchant Center.
- 2Review — you read the report. If parity isn't clean, the id mapping is adjusted and the analysis re-runs until it is.
- 3Cutover — the switch itself, gated on your explicit confirmation. Irreversible steps never run without it.
- 4Monitoring — after cutover, the feed guard and disapproval monitoring watch the new feed from the first submission.
The whole flow is resumable — if a phase is interrupted, it picks up where it stopped rather than starting over or leaving the migration half-done. And until you confirm the cutover, your existing provider keeps serving, untouched.
A feed migration should be boring. The parity report is how you buy the boredom in advance.