What this means for your store
A Hyper extract snapshots data locally for fast interactivity - ideal for leadership dashboards with millions of order lines. Live queries hit BigQuery or Snowflake on every drag-and-drop - slower but current. Intraday Google Ads pacing against today's revenue needs live or a short incremental extract; a weekly category margin review can run on last night's Hyper file without burning query budget.
Scenario on a real storefront
Wayfair's home category team runs two workbooks from the same BigQuery orders table - one for the trading desk, one for the executive SKU review:
// Executive SKU dashboard - Extract (nightly + incremental)
Data source: BQ orders → Extract
Incremental on order_updated_at
Refresh: 05:00, 12:00 UTC
Filters pushed: last 24 months
// Paid media pacing - Live (or 15-min extract)
Data source: BQ orders + intraday ads
Live connection
Custom SQL limited to CURRENT_DATE()
Viz: spend vs revenue by hour
// Cost control: live workbook gets row-level date filter in connectionWhat to do next
- Large GA4 export tables - always filter dates in the connection, extract or live.
- Incremental extracts need a reliable high-water mark column; avoid extracts on views that full-scan BigQuery each refresh.
- Show an “as of” timestamp on extract dashboards so media teams do not pace ads on six-hour-old revenue.
Bottom line
Extracts win speed and cost for historical ecommerce analysis; live wins freshness for intraday spend decisions. Mix both, label refresh times, and incrementalise big order tables.