The method

The factory is the product behind the product.

Nexcore runs its own operations on the same standardised loop it uses to build client systems. This page is that loop, written down in full — so you can judge the method before you buy the output, and hold us to it afterwards.

In one paragraph

The factory is Nexcore's repeatable build process for custom business software: map, spec, audit, build, verify, record — the same six steps on every slice of every project, no exceptions. Work ships in slices, one piece of working software at a time, scoped to named files, and a slice is finished only when it has been driven in a real browser against a live database, failure branches included — "it compiles" is not evidence. Each step leaves an artifact, so every build carries the same seven client-owned documents, from the workflow map to the verification report. The loop is deliberately boring: that is what makes the output predictable, why the record stays current, and why a second system costs less than the first.

Why a factory at all

Custom software usually fails in the same two places: nobody wrote down how the business actually works, and nobody checked the thing end to end before calling it finished. The loop exists to make both impossible to skip. It is boring on purpose — that is what makes the output predictable, and it is why the second system costs less than the first.

Every project carries the same six documents

Listed below. A new engineer — ours or yours — can pick up any project cold and know where things stand without asking anyone.

A slice isn't done until it's verified

Driven in a headless browser against the running system, across every route and both sides of every gate, with a screenshot attached to the report.

Nothing important lives in someone's head

Decisions, invariants and mistakes are written down as they happen — including a log of what we got wrong, so the same mistake isn't made twice.

Reuse before build

Auth, settings, notifications, invoicing and scheduling already exist as proven capabilities. Your build adapts them instead of re-deriving and re-testing them.

The loop

Six steps, every slice, every project

Each step below says what it produces, not just what it is called. A step with no artifact is a slogan, and you cannot hold anyone to a slogan.

  1. Map

    Before anything is designed, we write down how the work moves through your business today — the actors, the stages, the approval gates, and every exception path people currently handle by knowing to handle it. This includes the parts that only live in one person's head, which is usually where the real process is.

    Produces
    A workflow map: stages, actors, gates, branches.
    The point
    Most custom software fails because nobody wrote down the actual process — they built the one described in the meeting.
  2. Spec

    Work is scoped into slices. A slice is one shippable piece of working software with a defined edge, named down to the files it touches. Not a phase, not a sprint — a thing that either works end to end or is not done. You know what is in it and, just as importantly, what is explicitly not.

    Produces
    A scoped spec naming the files, the behaviour, and the exclusions.
    The point
    Scope drift happens in the gap between "build the quoting module" and knowing what that means.
  3. Audit

    Before a line is written, we run a read-only pass against the real codebase and the real database to find out what is already there. Findings are folded into the spec first. This is the cheapest step in the loop and it removes the most rework.

    Produces
    An audit note: what exists, what conflicts, what the spec got wrong.
    The point
    The expensive bugs are not typos. They are building something that contradicts what is already running.
  4. Build

    Implemented against a live database with real data shapes, never mocks or fixtures. Enum values, constraints and relationships are read from the schema rather than assumed, because a vocabulary invented in code and a vocabulary enforced in the database will disagree eventually — usually in production.

    Produces
    Working software against the real system.
    The point
    Something that works against fake data has not been shown to work.
  5. Verify

    Driven in an actual browser against the running system — every route it touches, at both mobile and desktop widths, and on both sides of every gate. The blocked path is exercised as carefully as the happy path, because the blocked path is the one protecting the business. Screenshots are attached to the report.

    Produces
    A test report with screenshots, per route and per viewport.
    The point
    "It compiles" and "the tests pass" are evidence of different things. Neither is evidence that a person can use it.
  6. Record

    Documentation is updated and committed as part of the slice, not afterwards. New invariants go in the playbook. Anything we got wrong goes in a miss-log with what we believed, what was true, and the tell that would have caught it sooner. If the docs are not updated, the slice is not done.

    Produces
    Updated living docs and a dated progress entry.
    The point
    Documentation written "later" is documentation that is never written.

The documents every build carries — yours, not ours

Written during the build and committed with it — which is the only way they end up accurate — then delivered branded, from your own portal. Everything here describes YOUR system and YOUR workflow. Our internal engineering notes stay ours; your documents are yours whether or not you ask.

System overview

The as-built operating manual of your system — every screen, every action, what happens on both sides of every gate.

Workflow map

Your business flow as captured in discovery — stages, actors, gates, branches.

Workflow flowchart

The visual of your workflow, branded to your business, with a deep-dive for every flow that branches.

Changelog

A plain-language record of every change shipped, dated.

What's next

The pending work, visible — not a promise in an email.

Release history

Every version, when it shipped, and what was in it.

Verification report

What was tested, on what, and the result — the "was it actually checked?" artifact.

What "done" means, precisely

A slice is done when all of the following are true. Not most of them.

  • Implemented against the live system, not mocks.
  • Driven in a real browser on every route it touches, at both mobile and desktop widths.
  • Both branches of every gate exercised — the blocked path as well as the happy path.
  • Core Web Vitals measured if the change touches rendering, and compared against the recorded baseline.
  • Living documentation updated and committed alongside the code.
  • Any new invariant recorded, and anything we got wrong logged with the tell that would have caught it sooner.

What the factory produces

Figures below come from build records — the test counts are real and were verified in a browser, not estimated.

683/683

Every test passing

on a live field-service platform, in production

611

Tests on a second build

a construction scheduling system

873+

Engineering hours logged

read from build records — not typed into this page · as of 28 Aug 2026

We publish figures we can point at a source for. Where we can't, we don't publish a number at all — which is why you won't find an ROI percentage or a time-saved claim anywhere on this site.