BigQuery to Snowflake migration without hand-written export jobs

Export or read the BigQuery tables, land them in cloud storage as Parquet, and load into Snowflake with COPY, or use a connector that does both steps and keeps them incremental. Run the two warehouses in parallel while you port queries, and compare row counts and column sums per table before you cut anything over.

SQL dialect, nested data and cost model. BigQuery's STRUCT and ARRAY columns have to become VARIANT, OBJECT or ARRAY in Snowflake, and any query using UNNEST needs rewriting against LATERAL FLATTEN. The cost model also inverts: you stop optimizing for bytes scanned and start optimizing for warehouse size and idle time.

STRING becomes VARCHAR, INT64 becomes NUMBER(38,0), FLOAT64 becomes FLOAT, NUMERIC and BIGNUMERIC become NUMBER with explicit precision, BYTES becomes BINARY, and BOOL becomes BOOLEAN. For time, BigQuery TIMESTAMP is an absolute instant in UTC so it maps to TIMESTAMP_TZ or a UTC TIMESTAMP_NTZ, while DATETIME carries no zone and maps to TIMESTAMP_NTZ. STRUCT, ARRAY and JSON all land as VARIANT.

Yes, though less than people fear for plain analytical queries. The rewrites cluster in three places: nested data access, date and time functions, and anything using BigQuery-specific syntax such as EXCEPT in a SELECT or table wildcards. Budget the effort by counting distinct query patterns, not total queries.

The BigQuery to Snowflake migration path from Adapters copies tables, casts BigQuery types to their Snowflake equivalents, and loads on an incremental schedule, so you can run both warehouses in parallel during a cutover instead of freezing reporting for a big-bang export. Field mapping takes about a minute and no code is required.

Yes. This BigQuery to Snowflake connector ships prebuilt: the field mapping is wired the moment you pick the pair, transforms are included, and the first sync can run within minutes. No code or engineering sprint required.

Pricing is flat and monthly: Starter at $49, Growth at $149, Scale at $399. Every plan includes this pair, visual field mapping, and per-record logs. There are no per-task or per-row fees, so the bill stays the same as volume grows.

Hourly on Starter, every 5 minutes on Growth, and down to every minute on Scale. Failed records retry automatically with backoff, and alerting plus a full per-record log come standard on every run.

No. Fields are auto-mapped the moment you pick the pair, and you can rewire any mapping visually before the first sync. BigQuery NUMERIC casts to Snowflake NUMBER(38,9) and BIGNUMERIC to NUMBER at its declared precision, STRUCT and ARRAY RECORD columns land as VARIANT so nested payloads survive the trip, TIMESTAMP stays UTC so no hour shifts on the way across, and STRING maps to VARCHAR. Incremental runs use an updated-at watermark per table, target tables cluster on the load date, and writes MERGE on the primary key so you can re-run a window during parallel validation without duplicating a single row.