Migration Approaches

There are multiple ways to build a data migration strategy, depending on an organization’s business needs. However, most strategies are classified into one of two categories:

Big Bang Migration

In the big bang approach data extraction, transformation, and migration are performed concurrently for all source instances. This approach is rational and highly effective when migration volumes are low and functional dependencies during loading are not a constraint. In such cases, a one-shot migration is performed immediately after going live or after major functionalities are deployed to production. It is preferable to have minimal or no delta migrations. Big bang migrations are planned over long weekends or offline hours, which works well for both greenfield and brownfield implementations. The various phases in this approach are typically iterative in nature and involve higher timelines during the planning, build and testing phases.

Criteria-based Migration

In the criteria-based approach, data is divided into multiple subsets or clusters based on predefined criteria. Source data can be grouped by source country instance, application criticality, usage, data groups (DG) 1, 2, and so on up to DG‘n’
This approach allows migration of a manageable cohort size over a period

Table 1: A comparison of the parameters of the big bang approach and criteria-based migration

Parameter	Big Bang	Criteria-Based
Volume	Suited for a small volume of data	Suited for a large volume of data
Reconciliation	Complex; only weekend window available for reconciliation	Simple; static, historical, live data to be reconciled independently
Downtime	Longer downtime window required during migration	Downtime required only for live data or predefined criteria
Operations and BAU	Impact is higher as downtime and reconciliation windows are small	Impact is comparatively lower
Stabilization	Comparatively lower effort; depends on extent of stabilization required	Procedural and manual updates might be required
Parallel run	Typically, not required if computational resources are assured	Required, making it comparatively more complicated
Rollback	Fallback approach is easy	Fallback approach needs planning
Overall risk	High	Low
Build Effort	High as the solution framework needs to be build specifically to cater to the migration requirement.	Low as the API driven approach is highly reusable and involves minimal refactoring of the existing APIs.

Table 2: Flavors of criteria-based migration

	Definition	Scenario	Benefits	Risks
Business process-driven(BPD)	Customer-centric approach which mimics the process when a customer places an order Reuses existing functional capability Complete overhaul of product, process, and systems	Business-driven transformation, greenfield scenarios	Direct impact on NPS Prioritizes impactful journeys Impacts all systems involved in the new stack, respecting the new business process Ensures data integrity across systems.	Change management Governance for convergent scenarios
Capability-driven(CPD)	Capabilities migrated by feature or sub-capability; realized by one or more systems An example of this approach would be to migrate the customers based on the Line of Business (LoB) or customer segment (B2C or B2B). The migration strategies adopted are also different as B2C migrations are customer centric, on-demand and responsiveness would be key.	Brownfield, IT-driven, business-centric approach	Establishes common vocabulary between business and IT Supports quick retirement of legacy applications	Commercial off-the-shelf (COTS) package might not be aligned with the definition of the CPD approach
Catalog-based(CLB)	New product categories or new offers in existing categories implemented on new stack Other products moved to new stack on stack stabilization An example of this approach could be to migrate offers in the postpaid mobile category first, followed by other categories like prepaid and mobile corporate offers.	Applicable to greenfield and brownfield Product-centric approach	Allows quick rollout of digital-native stack Facilitates gradual pull-based customer migration	Takes longer to retire legacy Managing customer experience across legacy and digital stacks