• SalesforceChaCha
  • Posts
  • ๐Ÿ’ƒ Weโ€™ve Been Moving Data for 20 Years. Is It Finally Time to Stop? ๐Ÿ•บ

๐Ÿ’ƒ Weโ€™ve Been Moving Data for 20 Years. Is It Finally Time to Stop? ๐Ÿ•บ

Every copy is a liability

Good morning, Salesforce Nerds! Iโ€™m gonna level with you โ€ฆ your data has a travel problem. ๐Ÿงณ

It leaves the operational database for the warehouse.

It leaves the warehouse for analytics.

Then a nightly job drags a slice into Salesforce, where it lands a day late and slightly different. ๐ŸŒ™

Every hop leaves a copy behind. Every copy adds the same bill: duplicate data, stale data, storage fees, pipeline upkeep, and one more place to govern access and retention. ๐Ÿ’ธ

Zero Copy Architecture asks an interesting question. What if the best integration is the one you never build? ๐Ÿค”

Salesforce Data 360 can read data where it lives. That sounds like magic, so letโ€™s find the catch. ๐Ÿ‘‡๏ธ

TABLE OF CONTENTS

๐Ÿ’ƒ Weโ€™ve Been Moving Data for 20 Years. Is It Finally Time to Stop? ๐Ÿ•บ

BORROW, DONโ€™T BUY

HOW ZERO COPY ACTUALLY WORKS

Zero Copy has two halves. Federation lets Data 360 read data that lives in an external platform. Sharing lets that platform read Data 360 data in return. ๐Ÿ”„

Federation comes in two flavors.

Query federation sends a query to the external engine, such as Snowflake, Databricks, BigQuery, or Redshift. The engine does the work and returns the rows. Data 360 processes them in memory and discards them. ๐Ÿ”

Filters and aggregations push down to the source, so you move answers instead of tables. You also pay the source for the compute.

File federation skips the engine. Data 360 reads Apache Iceberg tables straight from storage using its own compute. That cuts external compute fees and suits large datasets, but not every platform supports it, and it reads tables only, not views. ๐ŸงŠ

Salesforce recommends file federation wherever the platform supports it. Query federation works with every compatible system. ๐Ÿค

Sharing runs the other way. Your warehouse queries unified Data 360 profiles in place, so analysts get CRM context without an export job. ๐Ÿ“ค

FEDERATE, INGEST, OR BOTH

WHEN COPYING STILL WINS

 Not all data belongs in Zero Copy. Here is the trade.

Federation wins on freshness and footprint. Every query reads the source, so you get no sync lag, no second copy to store, and no pipeline to babysit. โšก

Ingestion wins on speed and processing. Local data answers fast and survives a source outage. Salesforceโ€™s architect guidance favors ingestion when you build a governed, unified customer profile. ๐Ÿ—๏ธ

Persistence matters too. What you build on top of federated data, such as unified profiles, segments, and insights, generally lands in Data 360 anyway. File federation also needs caching to support trigger-based features like data actions. ๐Ÿ””

Caching, called acceleration, is a copy by design. Use it for frequent queries or huge tables, and use live query for the occasional ask. ๐Ÿงฎ

Three questions settle most debates. Is the lake actively managed, fresh enough, and complete?

One โ€œnoโ€ favors ingestion. โœ…

Hybrid is common and healthy. Ingest the identity core, and federate the long tail of history and reference data.

READ THE FINE PRINT

THE TRADEOFFS NOBODY MENTIONS

Zero Copy doesnโ€™t mean zero processing, zero latency, or zero dependencies. The work moves. It doesnโ€™t vanish. ๐ŸŽญ

Two bills. Queries burn Data 360 credits, and the lake charges for compute and storage. A query-heavy workload can cost more than ingesting once, so run the math on both sides. ๐Ÿงพ

Borrowed performance. Your query runs at the speed of someone elseโ€™s system. A paused warehouse or a network hiccup becomes your segmentโ€™s problem. ๐ŸŒ

Pushdown decides everything. Here is the idea in conceptual SQL:

-- Good: the source filters and aggregates
SELECT Region__c, COUNT(*) FROM Orders__dlm
WHERE OrderDate__c >= '2026-10-01' GROUP BY Region__c;

-- Bad: drags a full table across the network
SELECT * FROM Orders__dlm;

Flattened types. Lakes often turn an Email field into a plain string. Data 360 loses context that downstream logic may need. ๐Ÿงฌ

Split ownership. Transformation logic spreads across ETL tools, the lake, and Data 360. When a number looks wrong, three teams point at each other. ๐Ÿ”ง

WHO OWNS THE TRUTH?

RETHINK THE INTEGRATION MAP

The old map ran one way: system of record, then warehouse, then every app that needed a copy. ๐Ÿ—บ๏ธ

Zero Copy redraws it. The lake stays home for governed data. Data 360 becomes a consumption layer that borrows what it needs and shares unified profiles back, with no outbound pipeline. ๐Ÿ”

Warehouses keep their job. They stop acting as a depot for Salesforce-bound extracts and go back to being the analytical home.

Governance moves to the source. Access controls, retention, and lineage live in one place instead of five. One copy to audit beats five to reconcile. ๐Ÿ”

That cuts both ways. Your lakeโ€™s quality now shows up directly in your CRM, and a sloppy lake can no longer hide behind nightly jobs. ๐Ÿชž

Ownership changes too. The team that runs the lake now owns an SLA for customer-facing Salesforce experiences. Write that down before go-live. ๐Ÿ“

ONE QUESTION PER PIPELINE

MOVE DATA ON PURPOSE

Zero Copy isnโ€™t a storage discount. Itโ€™s a better default question: what does this copy buy us? ๐ŸŽฏ

If the answer is speed, transformation, or persistence, copy the data and own it. If the answer is โ€œwe always have,โ€ stop. ๐Ÿ›‘

The goal isnโ€™t to eliminate data movement. Itโ€™s to stop moving data when the move delivers no business or architectural benefit. ๐Ÿ’ก

This quarter, list your five heaviest pipelines into Salesforce. Label each one federate, ingest, or hybrid. Then find the ones nobody can justify. ๐Ÿ”Ž

SOUL FOOD

Todayโ€™s Principle

"Everything should be made as simple as possible, but no simpler."

Albert Einstein

and now....Salesforce Memes

What did you think about today's newsletter?

Login or Subscribe to participate in polls.