Work
Client names are used with permission. The numbers are the ones the client tracked afterwards, not the ones that were forecast at kickoff.
How to read these
Every one of these had a named client, a real deadline and a number somebody was accountable for.
We have picked these four because each shows a different version of the same problem: something the organisation genuinely needed, that nobody had time or context to do properly.
We have changed nothing about how they turned out, including the parts that took longer than planned. Where a number is given, it is the one the client tracked after we left — not a forecast from the proposal.
Two years of accumulated integrations meant a single customer order took eleven minutes to fulfil and nobody could change it safely. We mapped every path first, then replaced the orchestration with one typed service — cutting fulfilment to under a minute without freezing feature work for a quarter.
11 min → 48 secThe team had been running experiments for three years and could not answer the only question that mattered: did any of it work? We instrumented the event layer properly, added the statistical rigour that was missing, and built the dashboard. Three experiments were quietly doing nothing at all.
+14% activationEvery product team had its own buttons. We designed and built the system, then spent more time on the migration path than on the components — because a design system nobody can adopt is a component library with better branding. Shipped in three months; still the single version in use today.
11 teams, 1 systemA previous supplier had left the codebase in a state nobody wanted to own. Two weeks of reading, a written assessment the client kept, and then a steady rebuild of the parts that were failing. Handover documentation was written as we went rather than reconstructed at the end.
Zero handover debtAcross all of them
80%
Come from a referral
Usually a previous client's colleague, which is the only marketing we do.
5 wks
Median time to first release
From kickoff to something running in front of real users.
100%
Code we handed over
Every commit, every credential, every pipeline. On every engagement.
0
Recurring retainers
Not a criticism of retainers. We would simply rather be judged on the work.
Afterwards
They pushed back on our first proposal, which nobody else had done. The second one was half the size and twice as likely to work.
The handover was the best part. We run the platform ourselves now and still have them on call for the things only they know.
Eleven people, all senior, all in the calls. That is not a staffing model, it is a decision about how work gets done.
Trusted by teams at
What we are after
Work that still makes sense two years after we leave.
No heroics, no single points of failure, no system that needs its author to keep running.
The engagements above started as a paragraph describing something that was not working. That is genuinely all it takes to begin.
Vi använder cookies för att förbättra din upplevelse, analysera trafik och visa personliga rekommendationer. Läs mer i vår integritetspolicy.
Nödvändiga
Krävs för att webbplatsen ska fungera (session, CSRF-token).
Analys & statistik
Sidvisningar, klick, sökningar — hjälper oss förbättra.
Marknadsföring
Personliga rekommendationer och kampanjer baserat på ditt beteende.