Learn
Migration tools by capability: what each kind actually does
Migration tools are sold by name and compared by demo. It is more useful to compare them by capability: what each kind translates, what it leaves, and what evidence it gives you. Four kinds cover the market: platform-native free tools, agentic converters, partner translation platforms and deterministic conversion. Most estates use more than one.
By capability
What does each kind of migration tool cover?
| Capability | Platform-native free tools (Lakebridge, SnowConvert AI, Microsoft’s SSIS routes) | Agentic converters (Databricks Genie Migrate, AI code assistants) | Partner translation platforms (Next Pathway, LeapLogic) | Deterministic conversion (Novuz) |
|---|---|---|---|---|
| SQL translation | Yes, to their own platform only | Yes; a model drafts each file, then checks it | Yes, across many sources and targets | Yes, by fixed rules, to five targets |
| Stored procedures | Varies by source: Lakebridge’s Morpheus for T-SQL; SnowConvert AI turns Teradata procedures into JavaScript procedures | File by file, as SQL | Listed by both vendors | Converted by rule; about 5% go to a resolver with engineer approval |
| ETL packages (SSIS) | Lakebridge lists SSIS as experimental; SnowConvert AI converts to dbt with limitations; Microsoft runs or rebuilds packages | Not listed | Listed by both vendors | Control flow, data flows and expressions converted by rule |
| Orchestration and schedules | Not converted for SQL Agent or BTEQ control flow; rebuilt on the platform’s scheduler | Not listed | LeapLogic outputs cloud-native orchestration | Rebuilt with the original dependencies |
| Built-in reconciliation | Lakebridge Reconcile compares named tables; SnowConvert AI has a data validation CLI | Syntax and intent checks; Databricks advises testing results on data | Next Pathway Tester; LeapLogic schema, record and aggregate validation | Every object run against the original on the same inputs |
| Evidence per converted object | Not stated | Not stated | Ask for a sample report | A reconciliation report for every object |
| Same output on a second run | Rule-based transpilers, yes; Lakebridge’s model-based Switch converter is experimental | Not guaranteed: output is generated | See vendor documentation | Yes; two runs diff empty |
| Price | Free | Bundled or per token | Not published | Fixed price after a $3,000 assessment |
Facts from each vendor’s published documentation, as on our options pages. Last checked: 2026-11-03. An object counts as automated only if it needs no manual edit, and a one-line fix counts as manual.
Reading the table
What the table says
- SQL translation is close to solved: every kind does it, and the free tools do it at no cost.
- The gaps are packages, schedules and proof. Those decide most of a migration’s effort and risk.
- Built-in reconciliation means different things: comparing tables after the fact is not the same as running each object against its original on the same inputs.
- Combinations work: a free tool for the SQL it handles well, and a tool that converts the remainder and reconciles every object, including the free tool’s output.
Next question
Where to read next
Which options exist for your source and target?
FAQ
Frequently asked questions
Which migration tools include built-in reconciliation?
Most include some check. Lakebridge Reconcile compares named source and target tables; SnowConvert AI has a data validation CLI; Next Pathway and LeapLogic offer validation products. Novuz runs every converted object against its original on the same inputs and reports pass, fail or review per object.
Which migration tools convert orchestration and schedules?
Few do it fully. The free platform tools do not list SQL Agent jobs or BTEQ control flow, so schedules are usually rebuilt by hand. LeapLogic outputs cloud-native orchestration, and Novuz rebuilds schedules with their original dependencies.
Which tools convert SSIS packages fully?
None converts every package without exceptions. Lakebridge lists SSIS as experimental, SnowConvert AI converts to dbt with limitations, and Microsoft runs or rebuilds packages in Fabric. Novuz converts control flow, data flows and expressions by rule; Script Task code needs review whichever tool you use.
Do agentic converters give the same output twice?
Not by design: a language model generates the output, so a second run can differ. They check syntax and intent, and Databricks advises comparing results on test data. That is why their output needs reconciliation before cutover.
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.