The standard advice for financial data integration is to build a central data warehouse. Pull everything from your accounting system, CRM, payment processors, and banks into one unified database. Single source of truth. One place to query everything.
This architecture terrifies me.
The Beginner's Logic
Centralizing data makes intuitive sense when you're starting out. Why query five different systems when you could query one? Reporting becomes simpler. Analytics get faster. You eliminate data inconsistencies by having one authoritative source.
The appeal of unified reporting is powerful. Finance team wants revenue numbers? Query the central database. Need to reconcile transactions across platforms? Everything's already joined in your data warehouse. Performance is better because you've optimized indexes for your specific queries.
This perspective assumes your central database remains available, accurate, and secure. Those are massive assumptions for financial data.
The Expert's Concern
System architects who've seen centralized databases fail understand the risks. You've created a single point of failure for your entire financial infrastructure. When that database goes down, every dependent system stops working.
Consider what happens during database corruption. A bad migration script runs. A synchronization bug writes incorrect data. Suddenly your authoritative source is wrong, and that incorrect data has propagated to every system querying it.
Distributed systems where each application maintains its own data store are more resilient. Your payment processor keeps working when your accounting system has issues. Bank reconciliation continues even if CRM is offline. Systems remain loosely coupled.
The Security Implications
Centralized financial databases are attractive targets. One breach exposes everything. Your transaction history, customer payment data, bank account information all sitting together.
Distributed systems contain breaches. Compromising your invoice database doesn't automatically expose payroll data. Each system can implement appropriate security levels for its sensitivity.
There's also the recovery problem. Restoring one corrupted database from backup is straightforward. But if your central warehouse contains data from a dozen sources with different update frequencies, what state do you restore to? You can't roll back just one component.
The Alternative Approach
Federated queries across distributed databases are slower but safer. You query each system as needed, accepting the performance cost for resilience. Each system remains the authority for its own data.
This feels inefficient until the day your central database fails and you realize you've built a house of cards.