We build and sustain corporate software.
Bespoke engineering and our own products, for processes that must be verified, not merely executed.
Four movements of the same work.
Consultancybefore
Design before building. Including what should not be built.
Developmentduring
Build what was designed, in deliveries that work.
Sustainmentafter
Keep it standing, against an environment that moves.
Analysiscycle
Read what the system returns in operation, and propose the adjustment.
And between one project and the next. Not every technology decision becomes a project: choosing a supplier, deciding between building and buying, reading the architecture someone else drew. Those are our work too, including when the system is not.
System, to us, has a definition.
Everyone calls what they deliver a system. Here the word carries five conditions.
A trail, not a screen
The system is the route, not the interface. Every step leaves a record of its passage, and the route can be retraced by someone who did not follow it.
Auditable by a third party
The trail can be checked from the outside, including by those who do not trust us, and without needing our permission.
Grounded in solid data
Nothing in the output exists without a traceable origin in verifiable data at the input.
The output is an input, not an end
The result is raw material for analysis. A system that produces a result and no measure of that result is incomplete.
Feedback
The analysis of the output becomes an assessment that guides the improvement of the process the system serves.
The system ends at the assessment.The decision is human.
We examine, classify and set down the assessment. Whoever decides is whoever answers for the decision.
How we build.
Five rules that precede the project. They do not change with the size of the client, with the deadline, nor with whoever wrote the code before.
The decision is recorded
Why the architecture is what it is, what was discarded, on what date. The cost of a system lies in changing it two years later, when nobody remembers why.
The record of a decision either exists, or it does not.
The machine is a tool
We use the tools of today, including those that learn. They speed up the work and they do not decide architecture, do not choose technology, do not sign off a delivery.
At every stage, a name answers.
No locked doors
Complete documentation, open standards, data that leaves whole. Changing supplier cannot cost redoing what has already been paid for.
An exit that depends on negotiation is not an exit.
Whoever builds, maintains
Software left alone does not stay as it was, it rots. Whoever answers the call is whoever wrote the system, not a team that has never seen inside it.
Sustainment has an owner and a deadline.
Silent failure is the worst defect
A loud failure is a problem. A silent one goes on producing effects while nobody sees. What we build records its own workings.
The origin of an error is located, not inferred.
The machine stacks the layers. The red is where someone needs to look.
Our systems classify, order and flag. They do not decide.
Whatever demands judgement is marked, and goes to a person who answers for it. Automating triage is a gain. Automating responsibility is not.
Where we stop.
We do not build what will not pay for itself.
A diagnosis that rules out its own project is delivered all the same.
We do not deliver a system only the specialist understands.
Whoever operates it needs to understand what the system does.
We do not accept a deadline that only closes with a shortcut.
A short deadline is met with a smaller scope, not with debt the client pays later.
We do not deliver a result without a trail.
If the conclusion cannot be reconstructed, it does not go out.
We do not build a system that decides on its own
in matters of legal consequence. We classify, order and flag. The decision belongs to whoever answers for it.
We do not push fashionable technology
when a simple tool will do.
