The airlines getting the best results from data aren't using Delta Sharing. They're doing something much simpler.
Here's what I keep hearing:
"Delta Sharing eliminates brittle ETL pipelines and overnight syncs. No more data engineering bottlenecks!"
Sounds amazing. Until you ask the follow-up question.
At one airline, the IT team was excited about eliminating their "nightmare ETL processes" with real-time data sharing. But when the revenue and finance teams started mapping where business rules would live, the room went quiet.
"Business rules?"
"You know — defining true OD. What's the logic for sales channels and sub-sales channels? Where do you apply distribution costs to PNR and segment level analytical products?"
More silence. The ETL wasn't just moving data. It was applying years of institutional knowledge, turning raw transactions into meaningful business information.
Delta Sharing can instantly share a booking record across systems. But it can't tell you whether that record represents new revenue, a change fee, or a refund without a cancellation — not without logic.
That logic has to live somewhere. And if it's not in your ETL, where is it?
Option 1: every consuming system rebuilds it — hello, inconsistency. Option 2: one team owns it all — hello, bottleneck. Option 3: it gets lost entirely — hello, wrong decisions.
Here's what vendors don't mention: storage is cheap, compute is expensive. These platforms make money when you constantly transform data, repeatedly applying the same business rules across multiple systems. Every query, every transformation, every "real-time" update drives their revenue.
Teams discover this during migration when their cloud bills explode — not from storing data, but from constantly reprocessing it.
We've been here before. Data Lakes promised to eliminate complex transformations. Data Mesh promised to push logic to domain teams. Now Delta Sharing promises to eliminate pipelines entirely.
But someone still has to define what "revenue" means. Someone still has to apply exchange rate logic. Someone still has to handle reference data for airport codes and fare classes.
The hype cycle sells the dream of eliminating complexity. The reality is you're just moving it somewhere else — usually scattered across departments where it becomes impossible to govern.
Here's the irony: your existing ETL already contains the solution. Those "brittle" pipelines document years of business decisions in code. Instead of eliminating them, successful organizations are mining them — extracting the business rules, documenting the logic, creating shared definitions that work regardless of the technology platform.
Making data work isn't about eliminating data engineering. It's about recognizing that your data engineers have been preserving institutional knowledge this whole time — and making sure that knowledge survives whatever comes next.