Data strategy and architecture
A data strategy is a plan that translates business strategy into data requirements. The architecture is what makes the plan buildable. I do both, together, because a strategy without an executable platform design is wishful thinking, and a platform built without a strategy is expensive activity.
What’s usually broken
- The strategy is a slide, and the slide is stale. There’s a vision deck from 2 years ago and nothing connecting it to what teams build this quarter.
- The architecture grew by accretion. One warehouse became 4, dev and prod share data, and nobody can draw the current state from memory.
- Every use case starts from scratch. No core reusable data assets, so each new request rebuilds the same joins with slightly different answers.
- Cloud posture is an accident. Multi-cloud by acquisition rather than by decision, with costs and access controls to match.
- AI ambitions, level-2 maturity. The board wants AI outcomes on top of foundations that can’t support them yet.
What I do
- Data strategy: vision and strategic pillars, a prioritized use-case portfolio (quick wins plus long-term bets, scored on impact and feasibility), maturity assessment, and an executional roadmap with owners and dates.
- Target-state architecture: platform design across ingestion, storage, transformation, and consumption. Medallion layering where it fits. Warehouse and lakehouse decisions made on workload evidence, including when open table formats earn their place and when they don’t.
- GCP and BigQuery specialization: platform consolidation on GCP, BigQuery warehouse design, and the surrounding services chosen deliberately rather than by default.
- Data and AI operating model: how the data function is organized, funded, and interfaced with the business. Centralized, federated, embedded, or hub-and-spoke, chosen for your company rather than by fashion.
- Multi-cloud strategy and build-vs-buy: written decision memos with real numbers, so the choice survives the meeting.
How it works
Strategy work starts with stakeholder interviews, because the strategy has to be written in the business’s language, and their words go in the final document. Architecture work starts from your actual estate: an inventory of what exists before any drawing of what should. The deliverable is a roadmap your team can execute without me, though most clients keep me around to help them do it.
Proof
At AXS I authored the multi-region target-state architecture, spanning US, UK/EU, and APAC tenants with a real-time branch and an agentic-systems layer, and the 5-phase rollout plan the organization adopted as its roadmap. At TelevisaUnivision I led the GCP-first consolidation and the move to a layered medallion architecture with contracts and quality validation at the seams. At Goldman, the Snowflake modernization that anchored the Marcus data platform.
Start a conversation
If you can’t draw your current architecture on a whiteboard, that’s a reasonable place to begin.