Data transformation

Turn raw source data into reliable decision infrastructure

SourceMedium applies cleaning, transformation, and enrichment so your warehouse and dashboards run on one coherent model layer.

  • Cleaning + transformation + enrichment
  • fct / dim / obt / rpt model shapes
  • 10 to 30% attribution coverage lift vs GA alone
Get started View modeling docs
Data transformation dashboard

Three layers of transformation quality

Cleaning, transformation, and enrichment each solve a different problem. Together they create stable output for analysis.

Data cleaning

Removes inconsistencies and standardizes source fields so downstream analysis is based on complete, usable data.

Data transformation

Converts raw source structures into warehouse-ready models that are easier to query, join, and report on.

Data enrichment

Adds context from trusted signals across systems to increase completeness for attribution and decision workflows.

Model architecture by use case

SourceMedium structures transformed data into model families so teams can choose the right level of abstraction for each workflow.

Model layer Purpose
Fact tables (fct_*) Immutable event or transaction records for metric calculations and trend analysis.
Dimension tables (dim_*) Descriptive context layers that support joins, segmentation, and standardized breakdowns.
One Big Tables (obt_*) Business-friendly semantic tables that simplify BI development and recurring analysis.
Report and summary tables (rpt_*, *summary*) Purpose-built and aggregated outputs optimized for dashboard and reporting use cases.

Quality and consistency controls

  • Schema alignment across source systems so equivalent fields map into one consistent structure.
  • Data type consistency for core identifiers and join keys to reduce downstream query failures.
  • Date and timezone normalization to avoid mixed UTC and local-time reporting errors.
  • Automated quality tests run multiple times per day to catch drift and invalid values quickly.

Practical outcomes

  • Consistent metric logic across dashboards, warehouse queries, and AI workflows.
  • 10 to 30 percent more attributable orders than GA-only coverage in many environments.
  • Lower analyst rework from definition drift and ad-hoc field cleanup.

Data Definitions

Canonical definitions for revenue, refund logic, channel mapping, UTM normalization, and deduping rules.

Attribution Source Hierarchy

Priority logic for selecting attribution signals when multiple sources disagree or are incomplete.

Data Cleaning Methods

Details on the cleaning layer, including standardization and automated validation workflows.

Data Enrichment Methods

How SourceMedium augments records with contextual signals and configurable business inputs.

Ready to stop debating the numbers?

Get started

Tell us a bit about your brand and stack—we’ll follow up shortly.