Home / Services / Business Systems / Custom Internal Platform
Custom Internal Platform
When the spreadsheet became a business system, this is the replacement.
The tool your business depends on has become the risk
Somewhere in the business there's a tool doing a job it was never designed for. A spreadsheet with years of growth and formulas nobody dares touch. Or a system built long ago by a developer nobody can reach. It started as a convenience. It is now the thing the operation actually runs on.
And it has started to frighten the people who understand it. Everyone who opens it can see everything, or change anything, and there's no record of who changed what. It has broken before, and work stopped while someone fixed it. New staff take months to learn its folklore. The person holding it together has become a single point of failure, and knows it.
What the team needs isn't database access. It's a real interface: screens built around the actual work, permissions so people see what they should, integrations so data stops being retyped between systems. That's a custom platform. It's serious money, which is exactly why I won't put a number on yours before doing the work of understanding it.
What you
actually get
Every line below is a deliverable, not an intention. If it's on this list it's in the proposal, and if it's not on this list it isn't included.
Paid discovery first, always
R45,000 – R75,000, fixed, agreed upfront. I sit inside the current system and the team that uses it, map the data, workflows, permissions and integrations, and find the requirements nobody thought to mention. This is not a sales exercise. It's the engineering that makes the build price honest.
A written technical specification you own outright
Data model, screens, roles, integrations, phase plan and a fixed price for the build. It's yours whatever you decide: take it to another developer, run a tender with it, or shelve it. It has standalone value, which is why it isn't free.
A build delivered in phases
Each phase ends with something real in use, starting with the core data and the screens a pilot group needs. No eighteen-month silence followed by a reveal. If priorities shift between phases, the plan shifts with them.
Role-based permissions
Each person sees and edits what their job requires, nothing else, and every change is attributable. The everyone-can-break-everything era ends at phase one.
Integrations with your existing stack
Accounting, email, whatever the spec identifies. Data entered once and flowing where it's needed, instead of retyped by the person who already has the most to do.
Documentation and a training programme
Written as the build progresses, not reverse-engineered at handover. Your team trained on their own screens, with their own data.
An agreed support period
Defined in the spec, in writing: what's covered, for how long, what happens after. Real use always finds things discovery didn't, and the support period exists to absorb them.
How it runs
So you know what's happening and when, and so nobody has to ask for a status update.
Discovery
Inside the business, with the system and the people who use it. Workshops, walkthroughs, and the awkward questions about workarounds. Paid, fixed and scoped before it starts.
The specification
The full written spec with a phase plan and a fixed build price. We walk through it, then you decide: build with me, take the spec elsewhere, or stop. All three are legitimate outcomes.
Core build
The data model and the first working screens, in the hands of a pilot group doing real work. Weeks to something usable, not months.
Permissions, integrations, the rest
Each phase priced from the spec, each one live and settled before the next begins. You can pause between phases. That's the point of phases.
Training and handover
The whole team trained, the documentation complete, and the old tool retired deliberately rather than abandoned.
Support period
As agreed in the spec. Fixes and tuning while real use finds the edges.
What I need from you
The projects that run late almost always run late for one of these reasons, so they're worth reading properly before we start.
- A decision-maker with real authority who answers within 48 hours, phased builds stall on absent decisions faster than on hard problems
- Access to the current system and the people who actually use it, especially whoever maintains it, they know where the bodies are buried
- A recorded walkthrough of the core workflows, and honesty about the undocumented workarounds, they decide whether the spec is right
- Hosting, infrastructure and licence costs are yours, in your name from day one, the spec includes a running-cost estimate so nothing surprises you
- A pilot group for phase one who will actually use it and complain properly, polite silence is useless to a build
What's not included
On the public page rather than buried in the proposal, so there's no argument in week five.
- A fixed price before discovery. Ever.
- Free scoping or specification work
- Hosting and infrastructure costs
What happens if you don't do this
Keep the current tool. Cheapest this year. But the risk compounds: every month adds data, dependencies and folklore, and the eventual failure gets more expensive. The real question isn't whether it breaks but what a bad week costs when it does, and whether the person holding it together will still be there.
Buy off-the-shelf. Sometimes the right answer, genuinely. If a product covers what discovery finds you actually need, I'll tell you to buy it, put that in writing in the spec, and the engagement ends there for the cost of discovery. That is the cheapest possible way to find out for certain.
Start smaller. Plenty of teams who think they need custom software need a properly modelled no-code database with automation and dashboards on top. That's The Operating System at R165,000 – R300,000, and if discovery points that way I'll say so, because a build at a fraction of the price that fits is a better outcome than a platform that flatters my invoice.
Straight
answers
Why won't you quote the build without paid discovery?
What happens to the specification if we don't go ahead?
You're one person. Who builds and maintains a platform this size?
Wouldn't an off-the-shelf tool be cheaper?
What does it cost to run once it's live?
Why does the build start at R450,000?
Sound like
your problem?
One call, no deck. If this isn't the right package I'll tell you which is.