English Vocabulary for Data Migration Projects

Data migrations are among the most stressful and high-stakes engineering projects. They combine technical complexity with business risk — and they require precise communication at every stage, from planning through to the post-migration review. Whether you are migrating a database, moving data between cloud providers, or consolidating schemas, the vocabulary is specialised and consistent across the industry.

This guide covers the English vocabulary you need to participate confidently in data migration planning meetings, write migration documentation, and communicate status during a live cutover.

---

What Is a Data Migration?

A data migration is the process of moving data from one storage system, format, or location to another. Migrations vary enormously in scope:

• Moving from one database engine to another (e.g., MySQL to PostgreSQL)
• Moving from on-premise infrastructure to a cloud provider
• Merging two databases after a company acquisition
• Changing a schema (adding columns, changing data types, restructuring tables)
• Moving data between regions for compliance reasons (e.g., GDPR data residency requirements)

What all migrations share is a need for careful planning, validation, and a clear rollback strategy.

---

Core Migration Vocabulary

ETL

ETL stands for Extract, Transform, Load — the three fundamental steps in moving data:

• Extract — reading data from the source system
• Transform — cleaning, converting, or restructuring the data to match the target schema
• Load — writing the transformed data into the target system

“We’re building an ETL pipeline to migrate the customer records from the legacy Oracle database into our new PostgreSQL schema.”

Variations include ELT (Extract, Load, Transform) where transformation happens after loading into the target.

Source and Target

• Source — the system you are migrating data from (“the source database”, “the source schema”)
• Target — the system you are migrating data to (“the target environment”, “the target table structure”)

These are the two most fundamental nouns in any migration discussion.

Schema

A schema is the structure or definition of a database — its tables, columns, data types, constraints, and relationships.

• Schema migration — a change to the database structure rather than the data itself
• Schema drift — when the actual database structure diverges from what is documented or expected

“Before we migrate the data, we need to run the schema migration to create the new tables in the target.”

Cutover

The cutover (also called cut-over) is the moment when traffic switches from the old system to the new one. It is the most critical moment in a migration.

• cutover window — the planned time slot for the switch
• hard cutover — an abrupt switch with a period of downtime
• rolling cutover — a gradual switch, often using feature flags or traffic routing
• zero-downtime migration — a migration strategy designed to avoid any service interruption

“The cutover window is 03:00–05:00 UTC on Saturday. If we’re not ready by 04:30, we abort and roll back.”

---

Data Validation Vocabulary

Validation is how you confirm the migrated data is correct and complete. It happens at multiple stages.

Reconciliation

Reconciliation means comparing data counts and values between the source and target to confirm they match.

• “Row count reconciliation: the source has 4,712,893 rows; the target has 4,712,893 rows. Counts match.”
• “We’re running a checksum reconciliation on the financial tables to ensure no values were altered during transformation.”

Data Quality Terms

• completeness — all expected records are present
• accuracy — values are correct and have not been corrupted during migration
• consistency — relationships between records are preserved (foreign keys, referential integrity)
• referential integrity — every foreign key in the target points to a valid record in the related table

“The validation suite checks for completeness, accuracy, and referential integrity before we sign off on the migration.”

Validation Phrases

• “Spot-check a sample of records” — manually verify a random selection to detect transformation errors
• “Run a diff against the source” — compare migrated data with source data
• “Validate the counts” — confirm record counts match between source and target
• “Flag anomalies” — identify unexpected values or patterns for review

---

Migration Strategy Vocabulary

Big Bang vs. Incremental

• Big bang migration — all data is moved in a single operation, typically with downtime. Higher risk, simpler execution.
• Incremental migration — data is moved in phases or batches, often while both systems run in parallel. Lower risk, more complex.

“We’re doing an incremental migration: historical data moves first, then we sync the delta before cutover.”

Delta and Backfill

• Delta — the new or changed data that accumulates after the initial migration while the source system is still active. “We need to sync the delta — approximately 80,000 records created since the initial load.”
• Backfill — populating a new column or table with computed values for existing records. “We need to backfill the new user_tier column based on historical purchase data.”

Dual Write

Dual write is a migration technique where new data is written to both the old and new systems simultaneously during a transition period, allowing you to switch the read path without a risky cutover.

“We implemented dual writes for 72 hours before cutover. Both databases stayed in sync, which gave us confidence to switch over with zero downtime.”

Shadow Mode

Shadow mode (or shadow traffic) means running the new system in parallel with the old one, processing the same requests but not serving the results to users. It lets you compare outputs without affecting production.

“We ran the new reporting database in shadow mode for a week, comparing query results with the legacy system. Discrepancy rate was below 0.01%.”

---

Rollback Vocabulary

Every migration needs a rollback plan — a defined way to return to the original system if something goes wrong.

• rollback plan — documented steps to revert to the source system
• rollback window — the period during which a rollback is feasible
• point-in-time restore — restoring a database to a snapshot taken at a specific moment
• abort criteria — the conditions that trigger an abort and rollback (“If the error rate exceeds 0.5% at cutover, we abort.”)
• dry run — a practice migration run in a non-production environment

“We have a rollback plan in place: we can restore the source database from a snapshot within 15 minutes if the cutover fails. The rollback window is 2 hours post-cutover.”

---

Key Vocabulary Summary

Term	Definition
ETL	Extract, Transform, Load — the three migration steps
Source	The system data is being moved from
Target	The system data is being moved to
Cutover	The moment traffic switches from old to new system
Reconciliation	Comparing source and target to verify correctness
Delta	New data created after the initial migration load
Backfill	Populating a new field for existing records
Dual write	Writing new data to both systems during transition
Shadow mode	Running the new system in parallel without serving results
Rollback plan	Documented steps to revert if migration fails
Referential integrity	All foreign keys point to valid related records
Dry run	A practice migration in a non-production environment

Data migrations are too important for vague communication. A single misunderstood term — “I thought the cutover was the rollback window” — can turn a recoverable situation into a disaster. Learn this vocabulary precisely, and you will be a clearer, more confident participant in every migration project you join.