The phrase "we need a system" arrives in our inbox in several forms. A distributor tired of tracking stock in Excel. A school managing fees on WhatsApp. A clinic whose appointments live in three diaries. What they need is a custom web application: software built around how their business actually runs, used through a browser, owned by them.

It is also the category of project most likely to go wrong, usually for reasons that were visible in week one. This is how the process looks when it is done properly, so you can recognise when it is not.

Stage one: discovery (one to three weeks)

Before anyone draws a screen, somebody has to understand the business. A good team spends the first stage asking irritating questions. Who enters this data? What happens when a payment is late? Which of these five reports do you actually open? The output is a written document: the users, the things they need to do, and the rules the system must enforce.

This is the stage that gets skipped when a client wants a quote "by Friday", and it is why so many systems are rebuilt eighteen months later. A number given without discovery is a guess dressed up as a price. Expect to pay for discovery on its own — typically Rs 50,000 to 250,000 depending on complexity — and expect the document to be yours to take elsewhere if you wish.

Stage two: design and prototype (two to four weeks)

Clickable screens with no real code behind them. You and your staff walk through placing an order, approving a leave request, whatever the system does, and you find the problems while they are still cheap. "The accounts team would never see it that way" costs a day to fix here and a month to fix after launch.

What to look for: are they designing for the person who will use it eight hours a day on a mid-range Android phone with patchy internet, or for the director who will look at it twice? The former is harder and matters more.

Stage three: the build (six weeks to six months)

The code gets written, usually in two-week cycles with something working to show at the end of each. The framework matters less than people think — we build on Laravel because it is mature, well-documented and easy for another developer to pick up later, which protects you — but a disciplined team on any modern stack beats a careless one on the fashionable one.

Three signs the build is healthy:

  • You can log into a staging copy and try things every fortnight, not just see screenshots.
  • The code lives in a repository you have access to, from day one.
  • Changes you ask for are written down with a cost and a reason, not absorbed silently and then presented as a bill at the end.

Stage four: testing and migration (two to four weeks)

The unglamorous stage that separates a system people use from one they quietly abandon. Real data from your spreadsheets gets imported and the mess inside it — duplicate customers, dates in four formats — gets reconciled. Your staff test with real tasks. Someone deliberately tries to break things.

Budget for this. Projects that "finish" without it are not finished.

Stage five: launch and the first ninety days

Training, a period of running the new system alongside the old one, and a flood of small requests once people use it in anger. A sensible contract includes a support window after launch, and a clear monthly rate for changes beyond it.

What it costs

Indicative ranges for Pakistan, for a team that follows the process above — not a quote, and the discovery stage is what turns these into one:

  • Rs 400,000 – 1,200,000: a focused internal tool — a booking system, an inventory tracker, a customer portal with a handful of roles. Two to four months.
  • Rs 1,200,000 – 4,000,000: a business operations system with several modules, approvals, reporting and integrations with payment or SMS providers. Four to eight months.
  • Rs 4,000,000 and up: platforms with many user types, high transaction volumes or regulatory requirements. Timelines are measured in quarters.

Offshore agencies quoting Pakistani clients in dollars tend to sit at two to four times these figures. Freelancers quoting a tenth of them are usually quoting for a prototype, whether they realise it or not.

Where projects in Pakistan go wrong

The fixed price before discovery. Both sides lose. The developer cuts corners to protect the margin; the client gets a system that does what was written, not what was needed.

The disappearing developer. A single freelancer with the only copy of the code is a single point of failure. Insist on repository access and documentation, however friendly the relationship.

Building what the owner wants rather than what the staff need. The owner approves the design; the staff abandon it in a month and go back to WhatsApp. Involve the people who will actually use it.

No budget for after. Software needs hosting, security updates and small changes forever. A system with no maintenance plan is a system with an expiry date — the maintenance checklist applies to applications even more than to websites.

Do you need custom at all?

Often not. If an off-the-shelf product does eighty percent of what you need and you can live with the twenty, buy it. Custom earns its cost when your process is genuinely different from everyone else's, when it is your competitive advantage, or when the subscription for the off-the-shelf tool would exceed the build cost within a couple of years. We talk clients out of custom builds regularly; it is better for everyone than a system nobody needed.

Frequently asked questions

Web application or mobile app?

A web application works on every device through a browser and costs a fraction to maintain. Build a native app only when you need something a browser cannot do — camera-heavy workflows, offline use, push notifications. The app-or-website decision guide walks through it.

Who owns the code?

You should, fully, once paid for. Read the contract. "Licence to use" is not ownership.

Can we start small and grow it?

That is the right way. Build the one module that hurts most, use it, then extend. A good architecture makes this cheap; discovery is where that architecture gets decided.

Our custom web application work always begins with a paid discovery stage that you own regardless of what happens next. Describe the problem — not the system you imagine, the problem — and we will tell you whether custom is the answer.