Capture · Zero-Copy Data Sources
Your warehouse data, live in Flash — without moving it.
Most integrations copy your data into another silo. Flash queries Snowflake, Databricks, Salesforce Data Cloud and Delta Sharing live, in place — governed by scope, a per-query cost budget, and a full audit trail. Your data never leaves your warehouse.
Audit entry
query.execute · scope · cost · ttl
The problem
Your richest data sits in a warehouse your marketing tools can't reach.
ETL pipelines that lag
A nightly export means your segments are always yesterday's data. Fast-moving signals — predicted churn, live spend — arrive too late to act on.
Copies that drift
Every copy is a liability: PII compliance, schema drift, stale joins. The more pipelines, the harder it is to know which number is right.
A wall between analytics and activation
Your data team built the models. Your marketing team can't use them without a custom data engineering sprint.
How it works
A governed query, not a pipeline.
Flash sends a structured query spec to your warehouse, reads the result into a TTL-bounded snapshot, then discards it. No ETL. No permanent copy. Every call is budgeted and audited.
No raw SQL
Queries are expressed as a structured FederatedQuerySpec — tables, filters, output columns. The dialect is compiled internally. No query injection path.
Cost-gated on every call
A per-query budget is reserved before the warehouse call goes out. Actual cost is settled against the reservation. Runaway queries are blocked, not throttled after the fact.
TTL, then gone
Results land in a time-bounded snapshot used for segmentation or personalisation. When the TTL expires, the snapshot is discarded — not archived.
Supported platforms
Connect the warehouse you already run.
Snowflake
Key-pair RS256 JWT
Compiled to Snowflake SQL. Connects via your private key — no password, no shared secret.
Databricks
OAuth M2M
Targets Unity Catalog tables. Client-credentials OAuth, scoped to the catalogs you specify.
Salesforce Data Cloud
Two-hop JWT bearer
Compiled to DC Query API v2. Connected user's JWT is exchanged for a DC access token on each call.
Delta Sharing
Bearer token (.share profile)
Open protocol — works with any Delta Sharing-compatible server. Predicate push-down via JSON.
Governance by design
Every query is scoped, budgeted, and logged.
Zero-copy doesn't mean zero-control. Every connection is scoped to the tables you allow. Every query runs under a cost ceiling. Every call writes an audit entry.
Scope control
Each connection specifies the exact tables and columns Flash is allowed to query. A query outside that scope is rejected at the guardrail, before it leaves.
Per-query cost budget
You set a maximum spend per query. Flash reserves the budget, executes, and settles the actual cost — blocking anything that would exceed the ceiling.
Full audit trail
Every query.execute event — connection, scope, cost, TTL — is written to the audit log. Retention, export, and access-control all apply.
What you can do with it
Live warehouse data, in the hands of your campaigns.
Zero-copy closes the gap between the models your data team builds and the campaigns your marketing team sends — without a pipeline in the middle.
Predicted-churn segments
Query your churn model output table live. Build a Flash segment from the result. Send a win-back campaign the same day, not next week.
Personalisation from the warehouse
Pull customer lifetime value, product affinity, or predicted next category from Snowflake into email merge tags — no manual export.
Marketing suppression lists
Suppress members who converted, opted out, or hit a service hold — using the freshest state from your warehouse, not a lagged copy.
Cross-channel attribution enrichment
Enrich Flash member records with offline or marketplace signals from your data lakehouse — one identity, all channels.
Why not ETL?
Zero-copy vs. a reverse ETL pipeline.
Data residency
Reverse ETL
A copy lives in the destination, under the destination's controls.
Flash Zero-Copy
No copy. Data stays in your warehouse. Flash holds only a TTL-bounded snapshot.
Latency
Reverse ETL
Hours to days — dependent on the sync schedule.
Flash Zero-Copy
Live — queried at segment build or personalisation time, not on a schedule.
Compliance surface
Reverse ETL
Every copy is in scope for PII compliance, access control, and data-residency rules.
Flash Zero-Copy
One system of record. No additional copies to govern, audit, or delete.
Engineering overhead
Reverse ETL
Connector maintenance, schema migration, backfill jobs.
Flash Zero-Copy
One connection configuration. No pipeline to operate.
What changes for the business
Your data team's models are in the warehouse. Your marketing team's campaigns are in Flash. Zero-copy closes the gap — without a data engineering sprint in between.
Live
warehouse data in segments — not yesterday's export
Zero ETL
no pipeline to build, operate, or break
Audited
every query scoped, budgeted, and logged