How to Write Requirements for a Custom Business System

Software strategy & delivery

Good requirements describe how the business works today and how it should work after the system is in place, step by step. They should cover workflows, roles and permissions, data, exceptions, integrations, acceptance criteria and non-functional needs such as performance, security and availability. Each requirement should be specific enough that the team building the system and the people signing it off would agree on whether it is met.

Why should requirements start with workflows?

Feature lists tell a developer what screens to build, but not how work moves through the business. Describe each process from its trigger to its end result: who starts it, what information is needed, which approvals apply, what gets recorded and who is notified. Use the real steps your staff follow, including the workarounds they use today.

Write the current process first, then the target process, and mark the differences. This shows which parts of the business are changing and which should stay the same, and it prevents the new system from quietly removing a step that people depend on.

Which roles and permissions should you define?

List every role that uses the system, including people who only view reports, approve requests or manage settings. For each role, describe the tasks it performs and the information it may see, create, change or delete. Include roles from outside your company, such as customers, suppliers or drivers, if they access the system.

Permissions are easier to get right when they are tied to the workflow. For example, a user who creates a purchase request may not be the one who approves it. Write these separation rules down, because they are often the difference between a system that is safe to use and one that needs manual checks.

What data does the system need to hold?

Describe the main data objects in plain language, such as customers, products, orders, shifts, invoices or assets. For each one, list the fields that matter, where the data comes from, who maintains it and how often it changes. Note which fields must be unique, which are required and which can stay empty.

Also describe the history you need to keep. Some businesses need a full change log for audit purposes, while others only need the current status. Clarify retention periods and any data protection obligations that apply to personal information.

How should exceptions and edge cases be written down?

Most requirement gaps appear in exceptions: a delivery returned after payment, a shift swapped at the last minute, a product code that does not exist in the master list. For each common process, ask what can go wrong, how often it happens and who decides what to do next.

Write each exception as a rule with its trigger, the expected system behaviour and the person responsible for resolving it. Exceptions that are not decided before development tend to be solved by hand later, which often means data is corrected outside the system.

Which integrations need requirements of their own?

For each external system, specify what data moves, in which direction, how often, and what happens when the transfer fails. Name the system of record for each data item, so the team knows which copy wins when two systems disagree. Integration requirements should also state how errors are reported and who fixes them.

As an example, an ERP project that synchronises online orders needs rules for matching products, handling cancelled orders and re-sending failed records. Our HMB ERP project shows how sales, warehouse and finance processes can be connected in this way, and the same questions apply to any integration.

What acceptance criteria and non-functional needs should you include?

Acceptance criteria describe how a requirement will be tested. Write them as observable outcomes, such as “a user without approval rights cannot submit a purchase order” or “a failed sync is listed on the exceptions screen with the reason.” Agree who will sign off each criterion before development begins.

Non-functional needs cover performance, availability, security, backup and recovery, supported browsers or devices, language support and accessibility. State them with measurable targets your business agrees to, and avoid vague wording such as “fast” or “always available” that cannot be tested.

Frequently asked questions

Do requirements have to be a formal document?

They need to be written and shared, but the format can be simple, such as a structured document or a spreadsheet. What matters is that each requirement can be read, discussed and tested.

How detailed should requirements be before design starts?

Detailed enough for the critical workflows, roles and integrations to be understood, with open questions listed. Less critical areas can be refined during design, as long as the open items have owners.

Who should sign off the requirements?

The process owners from the business should approve them, with IT or finance reviewing integration and security needs. A single named decision-maker helps resolve disagreements quickly.

Privacy policy