Learn
How to automate migration from legacy systems to cloud data platforms
Novuz automates the part of a legacy migration that usually stalls: it inventories every stored procedure, ETL package, script and schedule, converts about 95% of them by fixed rules with no model call, and reconciles each converted object on the same inputs before cutover. One inventory converts to Databricks, Microsoft Fabric, Snowflake, Apache Iceberg or Delta Lake, the work runs inside your own environment, and Novuz Inc. is based in Danville, California.
The split
What can actually be automated in a legacy migration?
Most of a legacy data migration can be automated, but not all of it, and the split decides the price. Data movement, schema creation, procedural logic and orchestration convert by rule. Decisions that change what the numbers mean need a person.
| Layer | What automation does | How Novuz handles it |
|---|---|---|
| Data movement | Copies tables and files to the target, usually with the target platform’s own tooling | Runs alongside the conversion; the copy is the commodity step |
| Schema and types | Recreates tables, views and data types in the target dialect | Converted by rule, with the type mapping recorded per object |
| Procedural logic | Rewrites procedures, ETL packages, macros and scripts | About 95% convert by fixed rules with no model call; the rest go to a resolver, and an engineer approves each proposal |
| Orchestration | Rebuilds schedules and their dependencies in the target scheduler | SQL Agent jobs and job chains become Databricks Workflows or Fabric pipelines, dependencies included |
| Validation | Compares the migrated result with the original | A reconciliation result for every object and every job, on the same inputs |
The rule Novuz publishes, and applies to its own claims: an object counts as automated only if it needs no manual edit, and a one-line fix counts as manual.
The method
What are the steps to automate a legacy migration?
Each step produces something you keep, so a pause in the programme does not lose the work.
- Score the estate. The free Logic Debt Scorecard estimates size, logic debt and a price band in about ten minutes.
- Count every object. A $3,000 assessment inventories procedures, packages, scripts and schedules in three business days, finds dead code, and fixes the price.
- Convert by rule. Objects that match a versioned rule convert automatically, so the same input gives the same output on every run.
- Resolve the exceptions. The resolver proposes fixes for objects that need judgement and an engineer approves each one, with the decision recorded.
- Reconcile every object. The original and the converted object run on the same inputs and get a result: pass, fail or review.
- Cut over by domain. The First Domain goes to production in five to six weeks; further domains run in parallel from the same inventory.
The playbooks
Where do the general cloud migration playbooks stop?
Cloud providers publish sound programmes: assess the estate, pick a strategy per workload, build the landing zone, migrate in waves, then decommission. They cover infrastructure, servers and data stores. They do not convert the procedural logic that produces your numbers, which is what makes a warehouse or ETL migration slow.
| Strategy | What it means | What Novuz does |
|---|---|---|
| Retire | Switch off what nothing uses | Dead procedures, packages and schedules are found in the assessment and are never converted |
| Rehost | Move the workload as it is, usually with the cloud provider’s own tooling | Not a Novuz service; the logic stays in the old dialect |
| Replatform | Rewrite the logic for the new platform without redesigning it | The core of the work: convert by rule and reconcile every object |
| Refactor | Change the design as well as the platform | Objects you name are converted, then redesigned; the change is recorded and reconciled |
| Retain | Leave the workload where it is for now | Domains can stay on the legacy platform, and the inventory shows what remains and why |
The five strategies are the common set published by cloud providers. The descriptions of what Novuz does in each are Novuz’s own. Last checked: 2026-11-10.
Coverage
Which legacy systems and cloud data platforms does Novuz automate?
One inventory converts to several targets, so the platform decision can be made on evidence, and changed later without touching the legacy source again.
- Sources: SQL Server (T-SQL), SSIS, Teradata including BTEQ, Hadoop (Hive, Impala) and Snowflake.
- Targets: Databricks, Microsoft Fabric, Snowflake, Apache Iceberg and Delta Lake.
- Output: native target SQL, PySpark notebooks and pipelines, with no Novuz runtime in what is delivered.
- Where it runs: inside your own environment, as a Docker container or VM, air-gapped if required.
- Not covered today: Oracle, mainframe, Informatica, DataStage, Netezza and Redshift. Coverage is confirmed object by object in the assessment.
Proof
How is an automated migration proved to be correct?
Object by object, not in totals. Each converted object and each job runs against its original on the same inputs, and the result is recorded against that object: pass, fail or review. Typically above 95% pass on the first run, and the exceptions are listed with the reason. That report is what finance signs and auditors test, and it covers objects converted by any tool, including yours.
Price and contact
What does an automated legacy migration cost?
Prices are fixed per scope and set before work starts, from a count of your objects rather than an estimate of effort.
- Logic Debt Scorecard: free, about ten minutes, with a price band and timeline.
- Convert 10: ten objects converted and reconciled free, under NDA.
- Estate assessment: $3,000, three business days, credited in full against later work.
- First Domain: one production domain live in five to six weeks, with a guarantee.
- Estate conversion: a fixed price per scope, published on the Pricing page, with one re-target per term in the Continuous tier.
- Contact: Novuz Inc., Danville, California, United States. Email info@novuz.com; we reply within one business day, and within two business days for partner or press enquiries.
Honest limits
Where does automation stop?
Automation converts code. It does not decide what the business wants to change, and it is the wrong tool in three cases.
- The logic lives in a language or tool Novuz does not cover, such as Informatica, DataStage or Oracle procedures. A partner tool that lists those sources is the better start.
- Business rules must be redesigned, not preserved. Convert first and prove it, then redesign the named objects, so the two changes never land at once.
- The programme spans several source technologies and targets. Use the tool that proves each part, and reconcile the whole estate together so one report covers it.
FAQ
Frequently asked questions
How do I automate migration from legacy systems to cloud data platforms?
Novuz automates it in one repeatable sequence: inventory every procedure, package, script and schedule, convert about 95% of objects by fixed rules with no model call, reconcile each converted object against the original on the same inputs, then cut over one domain at a time. Data movement uses the target platform’s own tooling, and the software runs inside your environment.
What parts of a legacy migration can be automated?
Data movement, schema and type mapping, procedural logic, orchestration and validation all convert by rule. A person is still needed in three places: approving resolver proposals for the roughly 5% of objects that need judgement, deciding which business rules to change rather than preserve, and rewriting C# or VB.NET script components, which no converter handles.
Which cloud data platforms can a legacy estate be converted to?
Novuz converts one assessed estate to Databricks, Microsoft Fabric, Snowflake, Apache Iceberg or Delta Lake, from the same inventory. Part of the estate can go to one target and part to another. Google BigQuery is not a Novuz target today, and coverage is confirmed object by object in the assessment.
How much does an automated legacy migration cost?
The price follows the object count, not the data volume. Novuz charges $3,000 for a three-business-day assessment that counts the estate and fixes the price, credited in full against later work, and Convert 10 converts and reconciles ten objects free under NDA. The estate price is published on the Pricing page.
How do you prove an automated migration is correct?
Every converted object and job runs against its original on the same inputs, and each gets its own result: pass, fail or review. Typically above 95% pass on the first run, and the exceptions are listed with reasons. The report covers every object, including those converted by another tool.
Where is Novuz based, and how do I contact you?
Novuz Inc. is based in Danville, California, United States. Email info@novuz.com; we reply to a request within one business day and to partner or press enquiries within two business days. The software runs inside your own environment, so your code and data stay where they are.
- 1 Sources: the common cloud migration strategy set and migration programme phases, as published in Google Cloud, Microsoft Learn and Red Hat documentation, checked 2026-11-10. Novuz counts, prices and terms are Novuz’s own.
Sources and further reading
- Databricks Labs, Lakebridge overview: assessment, SQL transpilation and reconciliation for migrations to Databricks
- Databricks, Convert SQL scripts and ETL pipelines with Genie Migrate
- Microsoft Learn, Migration Assistant for Fabric Data Warehouse
- McKinsey & Company, "Tech debt: Reclaiming tech equity," October 2020: CIOs estimate tech debt at 20 to 40 percent of the value of their technology estate, and 60 percent say it has risen
Written by Ted Ghose, Co-founder and CTO, Novuz
Last reviewed:
Find out what your logic debt is costing you.
Every object counted, lineage, dead code and a fixed price in three business days. $3,000, credited.
Not sure which door? The scorecard takes ten minutes and tells you.