TSA Exit Acceleration
Off the seller’s warehouse. On Newco analytics.
Fixed-fee proposal in 48 hours. Senior team on day one. The first conversation is always free.
TSA Snowflake Databricks separation is the work of standing up Newco data platforms on dedicated accounts with the right data slice, the rebuilt pipelines, the clean identity model, and the direct vendor commercial. The work runs inside the broader TSA exit strategy framework. The seller will not share its production data warehouse. Newco builds its own and migrates the data it needs.
The first step is a complete inventory of the data estate. Snowflake accounts, warehouses, databases, schemas, tables, and views inside each account. Databricks workspaces, jobs, notebooks, Unity Catalog metastores, and Delta Lake tables inside each workspace. Storage buckets that hold the underlying data files. Each artifact is classified as in scope for Newco, out of scope, or shared with seller residual entities.
The platform decision sits inside the inventory. Newco may stay on both Snowflake and Databricks. Newco may consolidate to one platform. Newco may evaluate alternatives such as Google BigQuery, AWS Redshift, or Azure Synapse for specific workloads. The decision rests on workload fit, integration footprint, total cost of ownership, and the engineering team strength. The decision is made before the program scope is finalized, not during execution.
Vendor commercial leverage is real for both Snowflake and Databricks at the carve out moment. Newco is a net new logo. The Newco team prepares committed credits, the BATNA, the term length proposal, and a feature scope. Pricing softens materially when the buyer pressure tests the proposal against named alternatives. The same discipline that applies to cloud separation applies to platform separation.
The TSA period covers the runway between Day One and the Newco production platform stand up. Newco continues to read from the seller platforms under a documented service catalog. The pricing follows the same cost-plus discipline that governs the rest of the TSA. The platform exit date is committed in the TSA exit ramp section, not left open.
The seller data warehouse holds the combined data of multiple business units. Newco does not need all of it. The data slice is the explicit definition of what moves to the Newco platform. Customer records that belong to Newco. Transaction records that belong to Newco. Product records that belong to Newco. Reference data that Newco needs to operate. The slice is defined by the business and validated by data engineering before any extraction begins.
For Snowflake, the migration approach uses native tooling. Snowflake Data Sharing creates a read replica of the seller scoped data inside the Newco Snowflake account during the TSA period. Snowflake Replication moves the data physically when the Newco account is ready for production. Snowflake account cloning is not available across organizations, which means the rebuild is a deliberate engineering exercise rather than a one click move.
For Databricks, the migration approach uses Delta Sharing, workspace export and import, and Unity Catalog metastore federation. The Delta Lake tables live in cloud storage that can be replicated or shared. The notebooks, jobs, and clusters get exported and imported. Unity Catalog permissions and lineage get rebuilt in the new metastore. The rebuild often surfaces undocumented dependencies that complicate the migration.
The slice decision drives the migration cost. A small slice migrates in weeks. A complete data warehouse copy can take months and may not be permitted under data protection rules. The discipline is to migrate what Newco needs to operate, not to copy what the seller had built. The work aligns with the carve out data separation framework.
Data pipelines feed the warehouse. Hundreds of pipelines in some sellers. Each pipeline reads from a source system, transforms the data, and lands it in the warehouse. The pipeline inventory is the parallel artifact to the data inventory. Each pipeline is classified as Newco scoped, seller scoped, or shared. The Newco scoped pipelines move to the Newco platform. The shared pipelines get unwound into two parallel pipelines.
Orchestration tooling moves with the pipelines. Airflow, Dagster, Prefect, or the cloud native equivalents. The directed acyclic graphs get exported, ported, and re registered against the new platform connections. The connection inventory covers source databases, source SaaS APIs, source file systems, target tables, target reports, and target downstream services. Each connection re binds to the Newco identity model.
Downstream consumers re point. Business intelligence dashboards in Tableau, Power BI, Looker, or ThoughtSpot move from the seller warehouse connection to the Newco warehouse connection. Reverse ETL pipelines that push enriched data to Salesforce, HubSpot, Marketo, or other operational systems re point at the new source. Machine learning pipelines that train against historical data re configure their feature stores and serving endpoints.
The integration boundary becomes the audit artifact. Every system that read from the seller warehouse and now reads from the Newco warehouse is documented. The boundary doubles as the operational handover document for the Newco data engineering team that owns the platform after the program closes.
Identity is rebuilt in the new accounts. The seller identity provider does not extend into the Newco accounts. Newco signs on with its own identity provider integration into Snowflake using SAML or SCIM. Newco signs on with its own identity provider integration into Databricks. Role hierarchies, group memberships, and privilege grants are designed from a clean canvas around the Newco operating model.
Data classification informs the access model. Personally identifiable information, financial reporting data, regulated health information, and confidential business information each receive a defined access policy. Snowflake row level security and column level masking align with the policy. Databricks Unity Catalog grants and dynamic views align with the policy. The data steward function reviews each grant before it goes live.
Service accounts that used to live in the seller identity tenant get rebuilt in the Newco identity tenant. Workflow accounts. Pipeline runner accounts. Application service accounts. Each account is rotated to a new credential issued by the Newco identity tenant. The seller cannot leave behind a service account that retains access to the Newco data plane after exit. The rotation discipline is absolute.
Audit logging is configured to a Newco managed log store. Snowflake account usage views, Databricks audit logs, and platform access logs export to a Newco controlled location. The retention policy meets the Newco internal audit and regulatory requirements. The discipline aligns with the broader Day One cybersecurity framework.
Cutover sequences the data, the pipelines, and the consumers in a defined order. The data lands in the Newco platform first under a defined freeze window. The pipelines re point and run in parallel against both platforms for a validation period. The consumers re point after the parallel period confirms data parity. The cutover runbook names every owner, every dependency, and every rollback option.
FinOps discipline is essential after the cutover. Snowflake credit consumption and Databricks DBU consumption are both highly elastic. A poorly configured workload can multiply consumption by 5 to 10 times the steady state target. The Newco team configures resource monitors, warehouse autoscaling rules, cluster policies, and query timeout policies before the consumers go live. The first month of consumption is benchmarked daily against the steady state target.
Stabilization runs for 30 to 60 days after cutover. Pipeline error rates, query latency, freshness service-level objectives, and access escalations get monitored daily. The acceptance criteria for the TSA exit certification include pipeline success rate above the agreed threshold, no critical access incidents in the trailing 30 days, and no unresolved data quality escalations from the business.
Programs typically run between $500K and $3M depending on data volume, pipeline count, downstream consumer count, and platform complexity. The discipline holds when the data slice is honest, the pipeline inventory is real, and the cutover plan is rehearsed. The program is delivered under a Fixed Fee + Portfolio Retainer engagement model through TSA exit acceleration.
Public cloud account moves, identity rebuild, and the FinOps discipline.
Read the article →Banking platform decoupling, account opening, and the cash management plan.
Read the article →The data protection boundary, the cross border rules, and the discipline that holds.
Read the article →The 90-day governance, IT, finance, HR and procurement separation plan we run on live carve-outs. Get the playbook plus the bi-weekly Day One Letter — short, signal-heavy, buyer-side.
No spam. Unsubscribe in one click. · Read the overview first →
Fixed-fee proposal in 48 hours. Senior team on day one. The first conversation is always free.
Seven buyer-side moves to exit a Transition Services Agreement on time and below budget. The mark-up, the extension-fee curve, exit sequencing, and the 11-month calendar.
One tactic, one benchmark, or one pattern from a recent buyer-side engagement. Short. Signal heavy. Free.
Subscribe to The Day One Letter →Scope the data split tightly. Every table you over-migrate and every pipeline you leave on the seller is a line on next year's TSA invoice.