Data architecture: Everything in its place
Most companies can't answer basic questions: how many customers switched provider or why.
Every action your brand takes leaves a data trail. When arranged properly, these trails create a living, organic data structure.
Following the data trail
Each interaction (browsing, in-store visits, emails, clicks) leaves valuable signals. Organised, they reveal patterns.
The problem: it's complicated
Disconnected channels everywhere. Analytics. Phone systems. Physical stores. Forms. Email. No technical standardisation across the board.
Data mesh theory
Data Mesh organises data by domains (sales, purchasing, post-sale, marketing, finance) where each team owns making sure their data is "connectable" under shared context patterns.
But we're already doing this... aren't we?
Implementing Data Mesh requires pyramid standardisation from the start: uniform patterns, fields, and processes. Nearly impossible in evolving companies.
Tools
Each platform has its own architecture. Data Mesh demands data transformers that convert heterogeneous information into a coherent whole.
How we do it (4 steps)
- Label and group your information: find field matches (e.g. "Customer" across sales and invoicing).
- Plan the transformation: define how data flows between areas and its lifecycle.
- Implement transformers: naming with domain suffixes and foreign keys.
- Worked example.
Detailed example
Customer "Juan Gómez" across 5 domains:
| Domain | System | Key |
|---|---|---|
| Sales | Online shop | VEN_ID = 12345 |
| Marketing | Mailing | MAR_ID (FK VEN_ID) |
| Support | Live chat | SUP_ID (FK VEN_ID) |
| Social media | RS_ID (FK VEN_ID) |
|
| Web analytics | Google Analytics | AW_ID (FK VEN_ID) |
Email is not recommended as a foreign key (B2B users typically have several).
Before and after
Before (Sales):
id_cliente_ventas: 12345
fname: JUAN
lname: GÓMEZ
user_email: juan.gomez@email.com
After:
VEN_ID: 12345
VEN_NAME: Juan
VEN_LASTNAME1: Gómez
VEN_EMAIL: juan.gomez@email.com
The reality
It's not straightforward. This is a theoretical framework and a solid foundation, but making it work in practice means dealing with tooling, configuration, and the odd headache.
CTAs
- Specialist course:
/en/services/guia-de-arquitectura-de-datos-en-empresas/ - "Book my session" → 30-minute networking call
