Change Data Capture
Also: CDC, Change-Data-Capture, Data Change Capture, capture des changements de données, capture de données modifiées, Erfassung von Datenänderungen, Änderungsdatenerfassung
A technique that detects and delivers only the data that changed in a source system, keeping downstream systems current without reloading everything.
What It Is
Change Data Capture (CDC) is a method for tracking every insert, update, and delete in a source database and passing only those changes to other systems. Instead of copying an entire table every night, CDC moves just what changed, often within seconds. A retail CFO sees the practical value when yesterday's sales appear in the finance dashboard by breakfast rather than a day late, because only the new transactions travelled, not the full ledger.
Why it matters
Most leaders never touch CDC directly, but they feel its absence. When a customer updates an address in one system and the change takes a day to reach billing, marketing, and support, the friction shows up as duplicate mailings, wrong invoices, and stale AI recommendations. CDC is what makes "real-time" a realistic promise rather than a slide, because it keeps warehouses, CDPs, and AI models fed with fresh data at a fraction of the cost of full reloads. For a CDO, it lowers the load on production systems while raising trust in reporting. For an AI track, models trained or prompted on data hours old behave differently than models on data days old, and CDC is often the deciding factor.
How it works
CDC usually reads the database transaction log, the same record the system keeps to recover from a crash. Reading the log means CDC captures changes without slowing down the applications people are using, unlike older methods that repeatedly scanned tables. Each captured change becomes an event: this row was updated, this one deleted, at this time. Those events flow through a data pipeline into a warehouse, lakehouse, or streaming platform. When you approve a project that promises "live dashboards" or "synced customer profiles across channels," CDC is very likely the mechanism underneath. Your practical checks are simple: how fresh does the data need to be, what does the source system tolerate, and who owns the pipeline when it breaks.