Services / Solution Architecture & Advisory

The most expensive mistakes in data and AI happen before any code is written.

A solution architecture engagement answers the questions that decide whether your investment pays off: what to build, what to buy, what order to do it in, and what to skip entirely. You get the thinking without committing to a build.

The full picture

Design backwards from the decision

Most architectures start from the data and the tools, then go hunting for a use case. The reliable approach runs the other way: start from the decisions the business needs to make, define the data products that serve them, then map the sources and systems that feed those products. Each data product gets one owner and a written service level, so the design survives contact with real operations.

Design first

Decisions & use cases

Three to five concrete decisions the business must make, each with an owner and a number it moves.

  • The question asked
  • The metric it moves
  • Who decides
defines

Design second

Data products

The smallest valuable datasets that serve those decisions, each treated as a product.

  • One accountable owner
  • Documented & discoverable
  • Service levels in writing
defines

Design last

Sources & systems

Only now: the systems, tools, and pipelines that feed the products. Technology fits the need, not the reverse.

  • Existing systems mapped
  • Build vs buy per layer
  • Team skills respected
Design runs this way: decision → product → sourceData flows back the other way once built

Approach informed by domain-driven design and Thoughtworks' published work on designing data products.

The main ideas

How we think about it

The principles behind the work, in plain language. If these make sense to you, we'll get along.

01

Fit over fashion

The right architecture is the one your team can run, not the one trending on conference stages. We design around your existing skills, systems, and budget, and we say so plainly when the boring option is the better one.

02

Total cost of ownership

Every component you add is something your team maintains forever. We weigh build versus buy, managed versus self-hosted, and the real long-run cost of each choice, not just the launch cost.

03

Designed for handover

An architecture only works if the people who inherit it understand it. Every engagement produces clear diagrams and written decisions, including the reasoning, so future hires can pick it up cold.

04

Honest risk assessment

Where could this fail? What happens when data volume doubles, a vendor changes pricing, or a key person leaves? Good architecture has answers written down before any of it happens.

What teams miss

Built for the demo, or built for the operating model

The difference between a pilot architecture and a production one is rarely the technology. It is whether anyone designed for the years after launch week.

The pilot architecture

  • Starts from the tools someone wants to try
  • The diagram gets drawn after the build, if at all
  • Hard-wired to a single use case
  • Works while the consultant is still in the building
  • Nobody owns the components after go-live
  • Scale, cost, and failure: 'we'll cross that bridge'

The production architecture

  • Starts from the decisions the business needs to make
  • Decisions and reasoning written down before code
  • Generalized across several use cases, so it gets reused
  • Designed for the team that operates it, at their skill level
  • Every component has one owner and a service level
  • Failure modes, growth, and exit costs assessed up front

In production

What a production-grade design covers

These are the artifacts and decisions a serious architecture engagement leaves behind. If a proposal does not mention them, ask why.

01

A use-case backlog

Three to five concrete decisions the platform must support, written with the owner of each. Designing against several use cases is what stops the architecture overfitting to the first one.

02

Data products, not just databases

Each dataset that matters is treated as a product: discoverable, documented, trustworthy, secure, and valuable on its own, with a single accountable owner.

03

Service levels in writing

How fresh must each dataset be, how long can the platform be down, who answers when it breaks. Agreed before build, so expectations are never a surprise.

04

Build vs buy, layer by layer

Ingestion, storage, transformation, serving, BI: each layer gets an explicit recommendation with the cost and the lock-in spelled out.

05

Security and governance early

Access control, PII handling, and audit requirements are design inputs, not a retrofit. Cheaper by far to design in than to bolt on.

06

An evolution path

How the design absorbs the next use case, the next tool, and ten times the data without a rewrite. Architectures are judged by their second year, not their first month.

The flow

How an engagement runs

A written, defensible plan your team can execute with us or without us.

01

Listen

We start with your business, not your tech stack: what decisions you need to make, what's slowing you down, what you've already tried.

02

Map

We audit what exists today: systems, data flows, team skills, and the constraints that are real versus assumed.

03

Design

We produce the architecture: diagrams, component choices, build-vs-buy calls, and a phased roadmap with rough costs.

04

Defend

We walk your team and leadership through it, take the hard questions, and revise. You keep everything, whoever builds it.

Sound like your situation?

A free 30-minute call to talk it through. No pitch, no obligation.

Book a call