A mobility platform's data team is overwhelmed by requests and every new dataset requires their involvement. What is the structural problem?

The structural problem

The data team is a service desk, and its capacity is the ceiling on the organisation's ability to use data.

Every dataset requires them to build a pipeline, every question requires them to write a query, and every change requires them to prioritise it. The team's headcount is the constraint, and adding people scales it linearly at best.

What the platform should provide instead

1. Self-service ingestion. A team can register a source and get it landed, with schema, quality checks and lineage applied by the platform — without a data engineer building a bespoke pipeline.

2. Ownership by the producing team. The team that owns the operational system owns its data product: its schema, its contract, its quality, its documentation. Centralising ownership of everyone's data in one team is what makes that team a bottleneck, and no amount of tooling fixes it.

3. A published contract per dataset, so consumers depend on an interface rather than on a table someone might refactor.

4. Discovery. A catalogue answering what exists, who owns it, what it means, how fresh it is and what it depends on. Without it, every question starts with an investigation.

5. Standard transformation tooling, so analysts build derived datasets themselves within guardrails.

The trade the decentralised model requires

Central standards, distributed ownership. The platform team owns the infrastructure, the standards, the guardrails and the catalogue; the domain teams own their data.

Without central standards this becomes fragmentation — every team with its own conventions, formats and definitions, and cross-domain analysis becomes impossible. Ownership decentralises; standards do not.

The prerequisite most organisations lack

Ownership per business entity. Which system is authoritative for customer, driver, trip, payment. Without it, several datasets describe the same thing differently and every consumer must choose — which is the dominant source of complexity in enterprise data estates, and no platform architecture fixes it.