Software Architecture Consulting.

Architecture is the set of decisions that are expensive to change later: how data is stored and separated between customers, how the parts of the system talk to each other, where work runs in the background, and how the whole thing is deployed. Get them right and new features are cheap. Get them wrong and every release risks breaking something unrelated.

Most products do not need a grand redesign. They need a handful of specific fixes, made in the right order, by someone who has run similar systems in production.

Pages slow down as data grows

Queries that were fine at a thousand rows time out at a million.

Every release breaks something unrelated

The codebase has no clear boundaries, so changes leak.

Cloud bills grow faster than revenue

Usually from over-provisioning, chatty services or uncached paid APIs.

You are adding customers, regions or a SaaS tier

Multi-tenancy, data isolation and localisation are hard to retrofit.

You are about to raise or sell

Investors and acquirers will run technical due diligence; better to know the findings first.

Your team is debating microservices

Before splitting anything, it pays to know whether the problem is really the architecture.

For most products under a few dozen engineers, a well-structured modular monolith is faster to build, cheaper to run and easier to debug than microservices. Split out a service when one part genuinely needs to scale, deploy or fail independently, not because the architecture diagram looks more modern.

We wrote a longer guide on this decision: Monolithic vs microservices, the 2026 decision guide.

1. 01Goals and constraintsA kickoff with founders and engineers: growth targets, pain points, budget and the deadlines that matter.
2. 02Read-only accessTo the repositories, database schema, infrastructure and monitoring. We never change production during a review.
3. 03Map the systemDiagrams at the context, container and component levels, using the C4 model, so everyone shares one picture.
4. 04MeasureSlow queries, error rates, deployment frequency, cloud spend and test coverage, so findings rest on numbers, not opinions.
5. 05Report and roadmapFindings ranked by risk and effort, with a sequenced plan. We use the pillars of the AWS Well-Architected Framework as a checklist, whatever your host.
6. 06WalkthroughA session with your team to go through every recommendation and answer questions. Implementation is optional.

These are decisions from systems we built and still operate, not slideware:

Single-database multi-tenancy with enforced scopes

Every company’s rows share tables, separated by a tenant id and a global scope, with tests that fail if a new model skips isolation (Odobase).

An outbox for side effects

Customer messages and third-party calls are recorded as events in the same transaction and relayed after commit, so a retry never sends twice (Odobase).

One posting engine for money

Only one module writes the general ledger, inside the same transaction as the document that caused it, and corrections reverse rather than edit (Agenso).

Shadow mode before cutover

A new subsystem runs beside the old one until integrity checks pass, then switches in one locked transaction (Agenso).

Budgets on paid APIs

Daily tallies and caching on every billable maps and data API, so usage is designed in rather than discovered on the bill.

A focused architecture review typically starts at about $3,000 and takes two to three weeks. System design for a new platform, or a review of a large multi-service system, is usually $6,000 to $15,000. The fee covers the report, diagrams and walkthrough; implementation is quoted separately if you want us to build it.