Vecture · Practice profile Enterprise data platform migrations Microsoft Fabric · Databricks Belgium · International

Enterprise data platform migrations.
As a specialist practice.

Data platform migration is its own discipline. Vecture is the practice of Metin Hatipogullari, who has run several end to end. The methodology and accelerator stack on this site are based on the lessons learnt from those engagements.

See the accelerators → Read the method
Where the platform is now.
Sources
Many stacks, loosely coexisting.
Oracle · SQL Server · SAP BW · Synapse
SAP BO · Qlik · SAS VA · Power BI · manual reporting
Shifting paradigms and people over a decade. Tribal knowledge, not documentation.
ETL · Transform
Logic the business runs on. Probably correct.
  • Built up over years of changing owners, paradigms and ways of working
  • SSIS, SAS EG, SAS DI, Informatica, SAP BW, stored procedures. Each with its own paradigm
  • Runtime queries (SQL, DAX, MDX) generated by the BI engine, never stored anywhere readable
  • Lineage, transformation logic, validation gates, DQ checks. All scattered across tool boundaries
  • Migrations stall here because nobody has a single coherent picture
Use cases
Different owners, evolving visions. Nobody knows for certain what’s critical.
SAP WebI · Qlik · SAS VA · Power BI · manual exports
Reports with data quality no one dares to question. Builds from teams that left years ago.
A reporting stack that keeps existing reporting alive, until it doesn’t. No room for new AI capabilities, increasingly difficult to maintain, hard to extend.
Vecture closes this gap
Where it has to land.
Any source
Connected, profiled, ingested.
Modern pragmatic ingestion. Full or incremental, optimised per source.
In migration context: temp data dumps untangle dependencies, turning serial pipeline chains into parallel workstreams.
Modern data transformation
Pattern-driven · governed · provably correct.
  • Standard open-source frameworks: PySpark, dbt, Spark SQL
  • Coverage from evidence. 18% of tables keeps 85% of reports alive
  • Three-way verification gate. No table promoted without proof
  • Maintainable in current and future widespread skillsets. Framework code and configuration
Any target lakehouse
Fabric · Databricks · or other.
Foundation for AI Foundry, data agents, and operational automation.
Not a reporting layer. A critical business asset.
The data platform the organisation acts on.
§ Accelerators

Proprietary instruments. Each adapts to the engagement.

Built from patterns across many past migrations. Selectively deployed* against the specific legacy stack and target platform. They accelerate the work without creating a dependency on proprietary code.

01discover
Source Harvesters.
SAP BO · SAS · SSIS · Informatica · stored procedures · extensible
Reaches the places metadata tools don’t. Runtime SQL from SAP BO and SAS. Package logic from SSIS. Procedure bodies from Oracle and SQL Server. Lineage graphs from Informatica repositories. A standardised model of what the estate uses, depends on, and queries. The foundation every other accelerator builds on.
Open →
02interpret
Legacy Data Miner.
Analytical pattern mining · coverage planner · join network · retirement scoring
Twenty-plus years of self-service BI produces thousands of reports. That is not a liability. It is signal. Data mining on actual analytical patterns reveals the data model the business runs on. 28 tables keeping 85% of 847 reports alive, derived from what the estate does rather than assumed.
Twenty-plus years of self-service BI produces thousands of reports. That is not a liability. It is signal. Data mining on actual analytical patterns reveals the data model the business runs on. The minimum table set that keeps the report estate functional, derived from what the estate does rather than assumed.
Open →
03interpret
Agentic Reverse Engineering.
Local & cloud LLMs · read-only legacy · cited findings
Agents with safe tool access read procedures, draft dbt models, flag Oracle-specific syntax with line citations. 340 procedures in 4 days, not six weeks. Human review at every gate.
Agents with safe tool access read procedures, draft dbt models, flag Oracle-specific syntax with line citations. Corpus-scale analysis in days, not weeks. Human review at every gate.
Open →
04verify
Migration Testing Workbench.
Business key comparison · source / legacy / target · column coverage · 18 quality checks
Three-way business key comparison across source, legacy, and target, from sprint two. Column coverage, row counts, SCD2 history, configurable per migration. An HTML quality report filed against every deployment, permanently. Cutover signed off on data.
Open →
05operate
Lineage-Gated Observability.
Quality + lineage + issue lifecycle · in development
Quality checks fire only when their upstream lineage is trustworthy. Issues are lifecycle objects with re-emergence detection. The DQ system the business actually acts on, because it stops crying wolf.
Open →

* Not a fixed product menu. Connectors, parsers, comparison rules get tuned per engagement, because the SAP BW dialect of one organisation is not the SAP BW dialect of another. New connector families extend a pattern already built. Methodology → for how they compose across the phases of a migration.

§ Engage the practice

Two main ways the work starts.

Pre-assessment1 wk to 1 month · scoped to migration size
Migration pre-assessment.
External read of the migration case before architecture and procurement commit. Patterns from past data platform migrations applied to the specific estate. What is likely to work, and what is not.
ForProgramme directors, CDOs, heads of data evaluating a migration.
You getA defensible scope and risk register. An architecture recommendation. An effort and timeline range grounded in evidence from the live estate, not in assumptions.
StartSend a current-state summary. Reply within a working day with a proposed next step.
Email →
Embedded deliverymulti-month · scoped
Embedded delivery.
Vecture embeds into the programme and works to one outcome: a successful migration. Role and reporting line are flexible. Lead, co-lead, or contribute in depth, whatever the engagement actually needs.
ForOrganisations committing to a full data platform migration.
You getThe accelerators deployed where they fit. Hands-on work through to cutover. IP transfer at close.
StartSend a programme brief. Reply within a working day with a proposed next step.
Email →

Also available: presales dossier reviews for SIs, consultancies and integrators in active pursuit (complimentary, up to 2 weeks). Triage sprints for migrations underway and slipping (1 week, fixed price). Email to discuss either.

§ Engage the practice

Get in touch.

info@vecture.be →