The honest answer nobody gives you

Ask five agencies what an app costs and you'll get five ranges, all of them vague, most of them ending in "it depends." It does depend — but not on magic. It depends on things you can actually see and control. Here is how the number is really built, in rand.

Real ranges for 2026

These are honest South African ballparks for custom, well-built software — not template resellers, not R500-on-a-freelance-marketplace:

  • A focused tool or automation (one job, done properly — a booking flow, a WhatsApp bot, an internal dashboard): roughly R3,500 – R25,000.
  • A full product MVP (auth, payments, a real database, an admin panel, a polished front end): roughly R45,000 – R180,000.
  • A custom platform (multiple user types, integrations, AI, ongoing scale): R180,000+, usually staged across milestones.

If a quote is wildly below these, ask what's being cut: security, testing, ownership of the code, or the person's time to actually finish it.

What actually drives the number

1. The number of "surfaces"

Every distinct screen a different person uses is a surface: the customer app, the admin panel, the back-office view. Each surface roughly doubles the design and testing work. The single biggest cost lever is how many surfaces you truly need on day one — versus what can wait for version two.

2. Integrations

Payments (PayFast, local and international cards), WhatsApp, email, accounting tools, courier APIs — each integration is a small project of its own. Two integrations is a feature. Eight is a platform.

3. AI, done properly

An AI feature that looks good in a demo is cheap. An AI feature that stays accurate as your prices, stock, and policies change — grounded in your real data, monitored for drift — is an engineering commitment. Budget for the second kind, because the first kind embarrasses you in month three.

4. Who owns it at the end

Cheaper quotes often keep you renting. Confirm you own the code and the accounts. Ownership costs a little more upfront and saves you the entire rebuild later.

How to spend less without buying worse

The best cost control isn't haggling on the day rate — it's scoping ruthlessly. Ship the smallest version that solves the real problem, put it in front of real users, and let what they actually do decide version two. We built this into how we quote: our project shop lets you assemble a system feature by feature and see the price move as you add or remove, before you commit a cent.

We also work deposit-first and fixed-price — you get an itemised number, not an hourly meter that rewards slowness. If it takes us longer than estimated, that's our risk, not your invoice.

The real question

"How much does an app cost" is the wrong first question. The right one is: what will this app earn or save me, and how soon? A R60,000 system that recovers R15,000 a month in lost bookings pays for itself in a quarter. A R15,000 system nobody uses costs R15,000. Scope for the outcome, and the price takes care of itself.