Hakkoda × Snowflake
Built on your stack.
Every enterprise arrives with an estate. Most of it works and all of it is load bearing. Here is the pattern, the method, and eleven builds running on it today.
EpicSAPWorkdayVeevaOracle
SailPointUiPathAlationSalesforceYours
Why projects stall.
Rarely the platform. Epic holds the clinical record, SAP the transaction, Workday the people. None of them can pause for eighteen months. A roadmap that opens by displacing one has picked a fight it did not need.
The other stall is quieter. Architecture gets drawn, then meets the security team. Tenant location, residency, approved tooling, identity model. Those answers should shape the design. Instead they redo it.
The estate stays. A governed layer meets it.
Snowflake becomes where the source systems meet. Each keeps its own job and its own boundary. The seam is where the engineering lives.
Stays in place
Seam
What we build
System of record
Epic Clarity and Caboodle
Clinical record, vendor intellectual property, its own licensing terms.
Governed layer
Dynamic ingestion into Snowflake
Metadata-driven pipelines through Fivetran, Matillion or Azure Data Factory, with Epic Systems intellectual property protected throughout.
System of record
SAP S/4 and Business Data Cloud
The transaction. Mid-migration more often than not.
Governed layer
Zero-copy share, semantic layer
No copy is made at all. Reporting continues through an S/4HANA migration instead of pausing for it.
Identity platform
SailPoint, Workday, existing policy
Who can see what, already decided and already audited.
Governed layer
Access built to policy in force
Role-based access, automated provisioning accelerators, tagging and masking. Security review becomes a check, not a negotiation.
Approved tooling
Tools security already approved
Automation platforms, orchestrators, catalogs. Anything that has passed review can go live now.
Governed layer
We connect to the approved tool
The build uses what is already approved. No swap-out request.
Every source system connects differently.
Using one connection method everywhere means picking it before you know what you are connecting to. Epic, SAP and Veeva each need something different. Sometimes the right answer is not to move the data at all.
Source
Method
Why
Epic Clarity and Caboodle
Fivetran Local Data Processing, Matillion, Azure Data Factory
High table counts and vendor intellectual property constraints. Wellstar, Beth Israel Lahey, Montefiore and VCU Health each landed on a different one.
SAP Business Data Cloud
Zero-copy data sharing
No duplication, no sync lag, no second copy to govern. Used at Aryzta.
Veeva Vault, CRM, Network
Native connectors and third-party shares
Regulated clinical data with an external research organization on the other side. Used at Ionis Pharmaceuticals.
Legacy Informatica and Netezza
dbt and Airflow migration frameworks
Sunset on a timeline the client controls, with cost savings calculated first. Used at Children's Hospital of Philadelphia.
Epic Cogito, queried directly
No new layer
Part of the deliverable at Children's Hospital of Philadelphia was a decision matrix for when not to route through the Trusted Data Layer.
Security posture is architecture.
Access model, residency, vendor licensing and audit each change the design. Answer them at the whiteboard or pay for them in remediation.
Constraint
Where
What it meant
Existing access policy
Caliber Collision
Workday data landed in Snowflake under the company's own security, compliance and access control rules.
Identity governance
Bank OZK
Snowflake joined to SailPoint over Private Link. Audit-proofing and onboarding automation.
Vendor licensing
Wellstar Health System
A thousand Epic Clarity tables ingested with Epic Systems intellectual property protected.
Platform standard
Shure
SAP S/4 modeled into Snowflake in place of BW4HANA. Role-based access and multi-factor authentication carried across.
Eleven builds, running in production.
All Hakkoda-delivered. Entries without a number carry no confirmed metric and are shown as structural proof only.
Wellstar Health System
Healthcare
Epic, Workday, UKG, ZocDoc and Medimpact into Snowflake on Azure through Fivetran Local Data Processing, with role-based access control.
1,000+
Epic Clarity tables ingested with Epic Systems intellectual property protected. Patient 360 across 400 locations.
Edwards Lifesciences
Medical Device
Fast Healthcare Interoperability Resources and HemoSphere telemetry off intraoperative devices into a Snowflake and dbt stack.
97%
Data efficiency gains. Five-minute pipelines, down from several hours.
Aryzta
Retail & CPG
SAP Business Data Cloud joined to Snowflake by zero-copy data sharing, with a semantic layer across operational and analytical systems.
4:1
Data growth to cost increase. Same-day SAP general ledger visibility across global regions.
Advocate Health
Healthcare
Epic Clarity and Caboodle through a metadata-driven ingestion framework on Fivetran High Volume Replication, replacing manual ingestion.
70%
Of analytics products now carry performance monitoring, error handling and data quality validation.
Centric Brands
Retail & CPG
SAP reporting mapped into Snowflake data products through staged domains, while an S/4HANA migration ran underneath.
514
Key performance indicators developed, 118 in production. 492 data objects in eight months against a twelve-month commitment.
US Foods
Supply Chain
Oracle Exadata and OBIEE, plus a supplier collaboration portal on Snowflake, Amazon Web Services and Sigma.
$10M
Direct earnings contribution from the supplier portal. 6,000 users moved off Oracle OBIEE.
Beth Israel Lahey Health
Healthcare
Epic Clarity and Caboodle into Snowflake with dynamic pipelines across Amazon Web Services and Azure, ingested by Matillion. Four people, three months.
1,000+
Clarity tables and 400+ Caboodle tables. Up to 500 users onboarded the following year.
MileOne
Automotive
CDK, eLead, Xtime, myKaarma, CallRevu, Salesforce Marketing Cloud and Google Analytics into one governed source of truth.
Structural
Seven previously disconnected dealership systems consolidated. Unified reporting across every rooftop.
Bank OZK
Financial Services
Snowflake integrated with SailPoint over Snowflake Private Link.
Structural
Audit-proofing and user onboarding automation. No numeric metric on file.
Ionis Pharmaceuticals
Pharma
Veeva Vault, CRM and Network with clinical research organization files and Snowflake third-party shares.
Structural
Automated provisioning and access management on Hakkoda accelerators. No percentage metric on file.
Medtronic
Medical Device
Snowflake Cortex Analyst with three connected semantic models, built to the client's enterprise security and architecture standards. First of two engagements at this client. The second is in section 06.
Directional
Category managers surface supplier insights directly. No quantified outcome on file.
Custom builds that integrate with what exists.
Each build is a module that integrates with what the client already runs. Two of the four below are not cleared for external use yet.
In production
A governed Snowflake layer beside the systems a client already runs.The eleven builds above, integrated with Epic, SAP, Workday, Veeva, Oracle and SailPoint.
In production
Connecting through tools security has already approved.When a client's security team has signed off on a platform, the build uses it. No exception request, no waiting.
Not cleared
Agents orchestrated from the client's own automation platform.Live at a Fortune 500 medical device manufacturer, orchestrated from the platform their security team had already approved, with Snowflake agents underneath. Pending attribution and disclosure confirmation.
Not cleared
Cross-region identity propagation.Sensitive source data held inside its own region, only summarized context crossing to the governed layer, every action running as the connected user. Documented, not yet publishable.
Three takeaways.
Start from what is already running.
Every build on this page began with the client's existing systems. Not one of them opened by asking the business to replace something.
Fit the connection to the system.
Epic, SAP and Veeva each connect differently, so we test the options for each one rather than using the same tool everywhere. At Children's Hospital of Philadelphia the right answer for some questions was to keep querying Epic directly and not move the data at all.
Use what security already approved.
A tool that has cleared the client's security review can go live now. Replacing it means starting that review again, which costs months.
Every client, architecture and metric on this page is drawn from the Hakkoda and IBM proof library. Entries marked Structural or Directional carry no confirmed number and are shown as structural proof only.
Every named client on this page is Hakkoda-delivered. Where IBM Consulting work is relevant to a conversation, it is cited separately and attributed explicitly. The two are never blended.
Items in section 06 marked Not cleared are documented internally and are not approved for external use.
Hakkoda, an IBM Company© IBM Corporation 2026