Case studies
Work documented in full, mistakes included
Most agency case studies are results with the reasoning removed. These are the opposite: the investigation, the decisions, what we rejected and why, and what we would do differently. Numbers are measured or absent — never estimated to look better.
A note on what these are
Both studies below document work on our own production site, not client engagements. Client work is covered by confidentiality until permission to publish is granted in writing, and we would rather show verifiable work on infrastructure we control than publish an anonymised story you have no way to check. Every measurement here can be reproduced against this site.
Website performance optimization
A seven-day-old homepage that every automated check reported as healthy. The cache header was the cause; the verification method was the reason nobody noticed.
- Caching
- Core Web Vitals
- Next.js
SEO architecture implementation
Replacing hand-written internal links with a generated graph, catching a config flag that silently discarded every redirect, and moving editorial rules into the build.
- Internal linking
- Technical SEO
- Structured data
How these are written
Each study follows the same structure: the problem as it presented itself, the investigation that established what was actually happening, the implementation step by step, the technical decisions with the alternatives we rejected, the trade-offs we accepted and what they cost, the measured outcome, and the lessons worth carrying forward.
The sections most agencies omit are the ones we consider load-bearing. Naming what was rejected shows the reasoning rather than the conclusion. Stating the trade-offs shows that a decision had a price. And documenting the mistakes — a link generator that starved the end of every cluster, an audit that measured the origin instead of the visitor — is what separates a write-up from a brochure.
Where a number could not be measured, it is absent rather than estimated. Where a question cannot be answered honestly, such as whether an incident affected rankings, the study says so instead of implying an answer.