Global Automotive Leader: a multi-domain platform for connected vehicle data
Vehicle telemetry, CRM records and sales signals sat in three systems that never met. Unolabs built a global automotive leader a multi-domain platform on Confluent event streaming and a governed lakehouse, with domain-owned quality SLAs. Customer, vehicle and product now resolve to one lineage-tracked view that operational teams act on directly.
A concise view of impact and engineering focus.
Client-estimated 35-45% faster analytics delivery
Trusted single view of customer, vehicle and product
AI-ready mobility and lifecycle analytics
Three views of one vehicle lifecycle that never met
Rapid operating model transformation across connected vehicles, eCommerce and aftersales left a global automotive leader with a fragmented data estate. Vehicle telemetry accumulated in telematics silos, customer records lived in CRM, and sales signals sat in commerce systems — three views of the same lifecycle that never met.
The practical cost was that cross-domain questions went unanswered in time to act on them. How vehicle behaviour relates to service demand. How a customer's digital journey connects to their ownership history. Both are answerable in principle; neither was answerable on the day somebody needed it.
For a business whose strategy depended on real-time operational intelligence — high-fidelity lifecycle analytics, proactive customer engagement — the silo structure was a strategic constraint rather than an inconvenience. It also compounded. Every connected vehicle added to the fleet widened the gap between the volume of signal arriving and the estate's ability to govern it, so waiting made the problem more expensive rather than clearer.
Connected-vehicle programmes generate data faster than most estates can govern it — the architecture question is not how to store telemetry, but how to join it, governed, to customer and product context.
Why event streaming, and why domains instead of one schema
We designed and engineered a multi-domain data platform unifying telematics, CRM and sales signals around three governed domains: customer, vehicle and product. The backbone is a real-time event-driven architecture built on Confluent, feeding an automated ingestion factory sized for high-velocity telematics and IoT telemetry. Streams land in a governed enterprise lakehouse.
Three design decisions carried the weight. Event streaming rather than batch, because vehicle and customer signals lose their value in hours rather than days. A multi-domain model rather than a single warehouse schema, because customer, vehicle and product lifecycles evolve at different speeds and answer to different teams.
And lineage as a first-class requirement rather than a documentation task, because a platform feeding operational decisions has to be able to show its working. Enterprise-wide metadata standards enforce that lineage from raw signal through to consumable data product. None of the three choices is free, and streaming in particular is a standing commitment taken here because perishable signals justified it.
- Siloed domain unification across customer, vehicle and product lifecycles
- Automated ingestion factory for high-velocity telematics and IoT telemetry
- Real-time event-driven architecture using Confluent and streaming analytics
- Governed multi-domain lakehouse providing high-availability decision support
- Enterprise-wide metadata standards ensuring auditable data lineage
Streaming-first is a cost and complexity commitment, justified here by the perishability of vehicle and customer signals — batch-first estates can support retrospective reporting, but not proactive engagement.
Domain-owned quality SLAs, platform-owned guardrails
The target state moved past traditional reporting into a 'Mobility Intelligence Layer' — a governed tier in which every connected-vehicle signal passes through an auditable pipeline before it reaches a decision-maker or a model. Nothing arrives at a decision unaccompanied by its provenance.
Federated governance gives each domain team ownership of its own quality SLAs, so accountability for data fitness sits with the people who understand what the data means. Platform-level guardrails hold security and metadata consistent across domains, which is the half of governance that must not vary by team. Splitting it this way keeps the central team out of arguments it cannot win.
The result is a single trusted view of customer, vehicle and product that operational teams act on directly, and that downstream systems build on without re-validating provenance case by case. Governance here is not a compliance layer bolted onto the platform. It is the property that makes the platform's outputs usable at all.
Domain-owned quality SLAs turn data quality from a central policing function into a distributed engineering responsibility — the platform enforces the contract; the domain owns the content.
The outcome: telemetry that pays for itself
The engagement delivered client-estimated 35–45% faster analytics delivery and a trusted, real-time single view of the enterprise ecosystem. That is the headline figure; the durable part sits underneath it.
More durably, the platform now serves as the reasoning foundation for autonomous mobility services and predictive maintenance agents. Because signals arrive governed, joined and lineage-tracked, downstream teams build on them directly rather than re-engineering ingestion for every new initiative. That is where the compounding return on a platform investment actually shows up: in the projects that never have to solve ingestion again.
Telemetry that was previously an exhaust stream became a direct commercial driver, powering lifecycle analytics, proactive customer engagement, and the service intelligence connecting vehicle health to dealer operations. The open cost is retention. High-velocity telemetry accumulates whether or not anyone queries it, and the platform forces that decision rather than making it. Retention gets harder to change the longer it is left.
What we'd flag: the 35–45% analytics-delivery improvement is a client-side estimate against their prior estate, not an independently instrumented benchmark — and high-velocity telemetry retention needs explicit cost and downsampling policies as the connected fleet grows.
Technical blueprint of the solution
This diagram visualises the core architectural pattern, data flows, and security boundaries implemented during this engagement.
What to carry into the next sprint
Takeaway
Commit to streaming only where the signal loses value in hours, not days.
Takeaway
Split domains by lifecycle speed and ownership, not by convenience of schema.
Takeaway
Set telemetry retention and downsampling policy before the fleet grows into it.
Frequently asked questions
- When is event streaming worth its cost over batch ingestion?
- When the signal perishes. Vehicle and customer signals lose their value in hours rather than days, which is what justified a Confluent-based event-driven architecture for this global automotive leader. Batch-first estates support retrospective reporting perfectly well; what they cannot support is proactive engagement. If the decisions you feed are made weekly, streaming buys latency nobody will spend.
- Why split customer, vehicle and product into separate domains?
- Because the three lifecycles evolve at different speeds and answer to different teams. A single warehouse schema forces them onto one change cadence, which means the slowest reviewer sets the pace for everyone. Governed domains let each team own its quality SLAs and its business logic, while enterprise-wide metadata standards keep lineage auditable across the joins between them.
- Is the 35–45% analytics improvement independently benchmarked?
- No. It is a client-side estimate measured against their own prior estate rather than an independently instrumented benchmark. That prior estate was heavily siloed, so a large share of the gain is manual joining work removed rather than raw platform speed. We are happy to take you through what the client measured, and what it would mean against your estate.
- What does a connected-vehicle platform cost as the fleet grows?
- More every year, unless retention is governed deliberately. High-velocity telemetry accumulates whether or not anyone queries it, so explicit cost and downsampling policies matter as much as the ingestion design does. We would treat retention tiers, aggregation windows and the archive path as first-class platform decisions rather than as something settled after a surprise invoice.
- Who is accountable for data quality in a federated model?
- The domain team that produces the data. Each domain owns its own quality SLAs, because accountability for fitness belongs with the people who understand what the data means. Platform-level guardrails keep security and metadata consistent across domains, so the split is clean: guardrails central, content local. Central policing of quality does not scale past a handful of domains.
Work out which vehicle signals actually need streaming
We will map your telemetry, CRM and sales flows against the decisions they feed, then tell you which need event streaming, which can stay batch, and where the governed join belongs.
Book a Streaming Assessment