
Data migration framework and services for South African organisations
Data migration is the planned move of records from one system to another, such as a database, ERP, CRM, mail tenant or cloud platform, without losing, corrupting or exposing them. Arke Technologies, in Johannesburg, runs data migrations for South African organisations to one documented framework: profile, map, cleanse, rehearse, reconcile and sign off.
What a data migration framework is, and why you need one
A data migration framework is the method that decides what moves, in what order, how it is transformed and how you prove it arrived intact. Tools move data; the framework makes the move governable. A data migrator, whether a person or a tool, working without one tends to copy old problems into the new system and find the gaps after go-live.
We apply it to a database transfer between servers, a system migration away from an old ERP, or mailboxes moving between tenants. It sits within our systems integration service and supports cloud migration projects.

Data migration definition and meaning: terms that get mixed up
The data migration definition is simple: moving data from a source system to a target system, once, and then retiring the source. Its meaning in practice is wider, because nearby terms describe different jobs.
| Term | What it means | Example |
|---|---|---|
| Data migration | A one-off move of data into a new system, with transformation and reconciliation | Customer, supplier and open-invoice records from an old ledger into Dynamics 365 |
| Database migration | Moving a database, its schema and its code to a new engine, version or server | Moving to a new SQL Server version, or from Oracle to Postgres |
| System migration | Replacing an application, with its data, users and integrations | An old CRM retired in favour of Salesforce |
| Data integration | A continuing exchange between live systems | A CRM creating invoices in Xero every day |
| ETL (extract, transform and load) | The technique behind both migration and integration | A nightly load into a data warehouse |
| Data porting | Moving data between services in a portable, standard format | Records exported to open formats so another service can import them |
Data migration vs data integration comes down to time: a migration ends; an integration runs as long as both systems do. Data migration vs database migration comes down to scope: a database migration can move data unchanged, while a data migration usually reshapes it. One search term to set aside: "migration data" in public statistics, such as the IRS migration data on household moves in the United States, describes population movement, not IT systems.

Our data migration process in eight steps
Our data migration process runs in eight steps, each ending in a document or a sign-off, so the project can be audited afterwards.
01.
Requirements for data migration
Migration data sources, owners, volumes, retention rules and the personal information involved; what moves, what is archived and what is deleted under POPIA.
02.
Profiling
Data quality in the source: duplicates, missing fields, invalid tax or registration numbers, orphaned records and inconsistent codes.
03.
Mapping and the approach document
A data migration approach document records every field mapping, transformation rule and exclusion, with a data migration flow diagram of sources, staging, target and load order.
04.
Cleansing
Your data owners fix what only they can decide; we automate the rest and log every change.
05.
Build
Extract, transform and load scripts or tools, kept in version control and repeatable end to end.
06.
Rehearsal
At least two trial runs into a copy of the target, timed and reconciled, until the numbers match.
07.
Cutover
A freeze window, the final load, reconciliation and a rollback point agreed in advance.
08.
Sign-off and retirement
Record counts, balances and sampled records signed off by your data owners, then the source archived or retired under your retention policy.
Data migration tools and software
We choose data migration software for each job, and often pair a platform's own tool with scripts for the transformations it cannot do.
| Job | Tools we assess | Notes |
|---|---|---|
| Databases to AWS | AWS Database Migration Service (AWS DMS) | Full load plus ongoing replication keeps the cutover window short |
| Databases to Azure | Azure Database Migration Service; Microsoft's SQL Server assessments, from the older Data Migration Assistant to the assessments in Azure's own migration tooling | We use the assessment to find compatibility problems before an Azure SQL Database migration |
| Databases to Google Cloud | Google Cloud's Database Migration Service | MySQL, PostgreSQL and other engines moving into Cloud SQL |
| Schema changes | Flyway and similar versioned-migration tools | Flyway data migration scripts make each change repeatable and reviewable |
| Microsoft 365 | Microsoft's migration tools; third-party tools such as BitTitan MigrationWiz | Mailboxes, OneDrive and SharePoint, including tenant-to-tenant moves |
| Google Workspace | The Google Workspace data migration service | Mail, calendars and contacts from other providers |
| File platforms | Vendor services and third-party tools | Box to Dropbox migration, Dropbox to SharePoint and similar moves, with folder permissions mapped |
| Salesforce | Data Loader and Data Import Wizard | The data migration tools in Salesforce, used after cleansing outside it |
| SAP | SAP S/4HANA Migration Cockpit; SAP Data Services (BODS) | Templates and staging tables for master data and open items |
Database migration: SQL Server, Oracle, MySQL, Postgres and NoSQL
A database migration process covers the schema, the data, the code inside the database (procedures, triggers and jobs) and the applications that connect to it. Missing the code is a common reason a database transfer appears to work while the application breaks.
| Move | What we do |
|---|---|
| Migrate a database to a new SQL Server | Assess compatibility, move logins and scheduled jobs as well as data, and test the application against the new version before cutover |
| Azure SQL Database migration | The same, plus checks for features the managed service does not support |
| Data migration in Oracle | Between versions or servers, with Oracle's own tools such as Data Pump |
| Data migration from Oracle to SQL Server, or to Postgres | Convert PL/SQL code and data types, tested object by object |
| Data migration in MySQL; Postgres data migration | Native dump-and-restore or replication, with checks on character sets and time zones |
| SQL to NoSQL migration, such as SQL to MongoDB migration | A redesign, not a copy: tables become documents shaped around how the application reads data |
Data migration to the cloud: AWS, Azure and Google Cloud
Data migration to the cloud is usually one workstream inside a wider move, so the strategy and wave plan sit with cloud migration. Database migration to the cloud needs its own tests, because a managed database's versions, extensions and maintenance windows belong to the provider.
| Target | Typical route | More |
|---|---|---|
| Data migration to AWS | AWS migration from on-premises to cloud often pairs server migration to cloud with AWS Database Migration Service, so a database replicates while people keep working | Amazon Web Services |
| Azure data migration | Azure Database Migration Service for SQL Server estates; Azure's storage tools for file shares | Microsoft Azure services |
| GCP data migration | The Database Migration Service in GCP moves MySQL and PostgreSQL into Cloud SQL | Google Cloud services |
| Cloud to cloud | To migrate from AWS to Google Cloud, or back, plan for egress charges, identity and network changes, and a period of running both | Cloud-to-cloud migration tools shorten the copy, not the planning |
ERP, CRM and SAP data migration
Business applications carry years of history, so the first decision is what not to move.
- SAP. Data migration with SAP covers SAP to SAP data migration and moves in from other ERPs. For SAP S/4HANA (S4 HANA) data migration, SAP's own data migration tools load master data and open items, and data migration in SAP BODS suits heavy transformation. Master data migration in SAP, such as material masters for SAP MM data migration, decides whether the new system works on day one. SAP silent data migration is different: SAP's own background conversion of application data after a technical upgrade. SAP's methodology supplies the SAP data migration framework; we deliver the extraction, cleansing, reconciliation and sign-off around it.
- Dynamics 365. D365 data migration uses configuration packages in Business Central and the data management framework in Finance and Supply Chain Management, loaded in dependency order and reconciled to the trial balance. See ERP implementation and support.
- Salesforce and Zoho CRM. Our data migration best practices for Salesforce: cleanse outside Salesforce, load parents before children, keep legacy IDs in external ID fields and pause automation during the load. Zoho CRM data migration follows the same order. See CRM implementation.
After go-live, the new system's links to everything else run through ERP and accounting integration and API development and integration; records that feed statutory filings keep their audit trail through SARS and CIPC integrations.

Microsoft 365, Google Workspace and reporting data
- Office 365 data migration. Mailboxes, OneDrive, SharePoint and Teams data move with Microsoft's or third-party tools; a tenant to tenant migration in Office 365 after a merger often uses BitTitan or similar, with domains, identities and licences cut over in a planned window. See Microsoft 365 licensing, migration and management.
- Google Workspace data migration. Moves into Google Workspace, called G Suite until 2020, use its data migration service for mail, calendars and contacts, with other tools for Drive; moves out of it use the target platform's tools. See Google Workspace licensing, migration and management.
- ETL, warehouses and Power BI. ETL in a data warehouse extracts data from source systems, transforms it into one consistent model and loads it on a schedule. ETL in Power BI happens in Power Query, which suits departmental models; larger estates move it into a data platform such as Microsoft Fabric or BigQuery. See Power BI consulting and dashboards.

Data migration key considerations and risks
Silent data loss. Truncated text, broken characters and shifted dates hide without reconciliation, so we compare counts, totals and checksums table by table.
Broken relationships. We load in dependency order and test referential integrity, so no invoice loses its customer.
Downtime. Replication or phased loads keep the freeze window short, and the cutover plan names who decides to roll back.
Personal information. POPIA applies throughout: test runs use masked data where possible, copies are encrypted, and temporary stores are deleted, with evidence, at the end.
Scope creep. Every exclusion is recorded in the approach document, so nobody discovers a missing year of history after go-live.
Access. Least-privilege accounts are created for the migration and closed when it ends.
Data migration cost and timelines
We quote each data migration project after discovery and publish no Arke rates. As a market reference, published South African rates for IT project work run from about R550 to R750 an hour ex VAT (matnet.co.za, September 2026). Cost follows the number of sources and objects, data quality (cleansing is often the largest task), transformation complexity, the history you keep, your tolerance for downtime and the number of rehearsals the risk justifies. A single database or mailbox estate can often move in days or weeks; an ERP or multi-system migration usually runs over months, paced by rehearsals and your data owners' sign-off.
Last reviewed: September 2026.

Governance, compliance, procurement and continuity
Governance and reporting. A named migration lead, a steering review at each phase gate, a risk and issue log, change control and a sign-off pack for every load.
Compliance. POPIA governs what moves: only the data you need, retention rules applied, and an operator agreement where we process personal information for you. Controls are aligned with ISO/IEC 27001; for public bodies the sign-off packs form the audit trail PFMA and MFMA audits test, and support King IV Principle 12 reporting.
Procurement. Level 1 B-BBEE contributor (sworn affidavit available on request), giving buyers 135% procurement recognition. Company documents, CSD supplier number and tax compliance status are available on request, and we complete SBD forms when a bid requires them. Buy through a proposal and purchase order, or through a tender response; see company facts.
Continuity. You own your data, credentials and documentation. Mapping specifications, scripts and reconciliation evidence are handed over at the end, and the source is archived under your retention policy.
Frequently asked questions
What is data migration? Show answer Hide answer
Data migration is the planned, one-off move of data from a source system to a target system, followed by retiring the source. It includes extracting the data, cleansing and transforming it to fit the new system, loading it and reconciling the result. A migration succeeds when your data owners can prove, with counts, balances and samples, that everything that should have moved did.
What is a data migration framework? Show answer Hide answer
A data migration framework is the repeatable method a team uses to plan, run and prove a migration: requirements, profiling, mapping, cleansing, build, rehearsal, cutover and sign-off. It fixes who decides what, which documents exist and how success is measured. Tools do the copying; the framework makes the migration auditable and stops surprises reaching go-live.
What is the difference between data migration and database migration? Show answer Hide answer
A database migration moves a database, with its schema and code, to a new engine, version or server, often without changing the data. A data migration moves data into a new application and usually reshapes it to fit, such as legacy customer records mapped into an ERP. Many projects involve both, so we plan them together.
What tools are used for data migration? Show answer Hide answer
Common data migration tools include AWS Database Migration Service, Azure Database Migration Service, Google Cloud's Database Migration Service, Salesforce Data Loader, SAP's Migration Cockpit and Microsoft 365 migration tools. Most projects also need scripts for transformations a tool cannot do. We choose tools per job and keep every script in version control, so each load is repeatable.
How long does data migration take? Show answer Hide answer
Data migration takes from days for a single, clean database or mailbox estate to months for an ERP or multi-system migration. The duration depends on data quality, volumes, the number of rehearsals and how quickly your data owners can sign off. We give a dated plan after profiling a sample, because profiling reveals most of the cleansing work.
How much does a data migration cost? Show answer Hide answer
A data migration is quoted per project after we profile a sample of your data, because cleansing effort, the number of sources and objects, the history you keep and your tolerance for downtime drive the cost. As a market reference, published South African rates for IT project work run from about R550 to R750 an hour ex VAT (matnet.co.za, September 2026).
Does POPIA apply to a data migration? Show answer Hide answer
Yes, POPIA applies to a data migration because moving personal information is processing it. You remain the responsible party, and anyone migrating the data for you acts as an operator under a written agreement. In practice that means moving only what you need, masking test data where possible, encrypting copies and deleting temporary stores, with evidence, when the project ends.
Plan your migration with one accountable team
Tell us what is moving, from where and by when, and we will profile a sample, map the risks and give you a dated migration plan. You can also reach us through our contact page or on 010 447 2445.