DESIGN STUDIES

Working through hard problems in commerce at scale.

Each study takes one problem, sets out the options, and explains the reasoning behind the choice — including what the choice costs.

Every study opens with a short, plain summary of what the decision gives you and what it puts at risk. The detail below it is for people who build these systems.

Requirements as a design input

Specification
What a good specification settles

A specification's job is to close the decisions the build would otherwise have to guess at. The ones worth settling are rarely on the happy path — they are the edges, and settling them early keeps each choice with the people whose choice it is.

Read the study →
Governance of shared practice

Consistency
Using AI as a consistency engine for architecture work

Most stories about AI in engineering are about writing code faster. The more useful question for an architect is whether it can make the thinking more consistent without taking over the decisions. Fix the structure, keep the judgement.

Read the study →
Decomposition & boundary design

Seams
Where the boundaries go in a multi-brand platform

One platform serving many brands and markets depends on where you draw the boundaries. Put the boundary where ownership and speed of change differ. Shared, stable behaviour on one side. Brand and market differences on the other, written as configuration and not as copies.

Read the study →
Consistency & coupling trade-offs

Events
Handling order events after checkout, and the one thing that cannot wait

Connecting every downstream system directly into checkout makes checkout fragile. Recording the order once and letting each system pick it up on its own time removes that risk — as long as you are honest about the one step that must still happen while the shopper waits.

Read the study →
Repeatability as an architectural property

Drift
Setting up a large storefront estate without drift

When sites are set up by hand, no two end up the same. Making the configuration the single source of truth — and making the step that applies it safe to repeat — turns setup from weeks into days and stops the differences at the source.

Read the study →