Security is often the strongest reason to modernise an established system. An upgrade that addresses recurring vulnerabilities can improve robustness and reduce the cost of investigation, maintenance and disruption. The decision still needs a clear business case: large codebases take substantial investment to replace, and a reliable targeted fix may deliver the required benefit much more economically.
A rewrite can be the right decision. Its scope and migration approach should follow the problem being solved, the needs of the users and the investment available. These seven questions help establish that direction.
1. Can the problem be solved reliably with targeted fixes?
Start with the most pragmatic option. If a vulnerability can be addressed with a reliable, well-tested change to the existing system, that is usually a better use of time and money than migration. A security issue does not automatically justify replacing an entire application.
Repeated classes of vulnerability can make a stronger case for modernisation. In ATM software, for example, persistent pointer or memory-management errors could justify moving affected components to a memory-safe language, where the platform and operational requirements permit it. That targets an underlying source of risk rather than repeatedly repairing individual defects. Investigate why related vulnerabilities recur, including weaknesses in design, testing or the development process.
Memory safety addresses a particular class of vulnerability. Authentication, permissions and business logic still need secure design and verification. A language should be chosen for the protections and capabilities it provides, with hardware access, timing requirements and interoperability considered alongside them. CISA's guidance on memory-safe roadmaps provides further context for this approach.
2. Is there a sufficient reason and investment to migrate?
An established system with limited complexity, dependable operation and little need for bug fixes may be serving its purpose perfectly well. Maintaining it can be a sound engineering decision. Its age alone does not establish that replacement will repay the cost.
Dependable operation is not evidence of security on its own. A system may produce the expected results while relying on unsupported dependencies or remaining exposed to newly discovered vulnerabilities. Retaining it should include reviewing support status, security updates and exposure as threats change. NCSC guidance on obsolete products explains the risks of continuing to use unsupported technology.
Migration needs a worthwhile objective and enough investment to achieve it. Account for understanding requirements, implementation, integration, testing, deployment and ongoing support. A budget that covers writing the replacement but leaves little room to establish confidence in it is an incomplete budget.
Where substantial investment is available, modernisation creates an opportunity to improve an ageing system's security, maintainability and user experience together. Funding makes the programme feasible; the expected benefits must still justify it. Make those intended improvements explicit so that the value of the work can be assessed.
3. What do users value, and what should improve?
Users should be a primary source of requirements. Find out what they like about the current system, what frustrates them and where they have developed workarounds. Those conversations establish which familiar capabilities are worth retaining and where a replacement can improve the experience.
Observe how people complete real tasks as well as asking about their preferences. Workarounds and support requests can expose difficulties that experienced users have learned to tolerate. Include support staff and users with accessibility needs, and test proposed improvements with them. GOV.UK's user-research guidance describes this approach to understanding users and their needs.
“You've got to start with the customer experience and work backwards to the technology.”
That principle is useful when modernising operational software. Establish the improvements users should experience, then choose technology and architecture that support them. Security and robustness requirements belong in the same discussion, including protections that users may never directly see.
4. Which existing behaviours are requirements?
Understand the current system thoroughly so that important constraints are identified. Then challenge its inherited design decisions. Separate the outcomes that must be preserved from the implementation choices that produced them, allowing new approaches to be evaluated against the actual requirements.
When reproducing existing functionality, detailed knowledge helps prevent incorrect behaviour. When changing the product or architecture, the same knowledge should inform the work without unnecessarily restricting it. Re-examine requirements with users and the purpose of the upgrade in mind.
Identify the interfaces, timings, data and operational workflows that genuinely need continuity. Existing bugs, awkward workarounds and obsolete assumptions should each be evaluated before becoming requirements for the new system.
5. Does the architecture need to change?
Be clear about the core reason for the upgrade. If the goal is a more secure and robust system, assess whether the existing architecture can deliver that goal. An architectural change may be necessary to meet the requirements.
Where the architecture remains suitable, improving individual components can preserve a useful foundation. Where that foundation is the constraint, a full rewrite or a new implementation may be more appropriate.
A new architecture can still reuse suitable existing components. Proven USB handling or socket implementations, for example, may support the new system if their security, behaviour and interfaces remain appropriate. This allows useful existing work to support a new direction.
6. How can old and new work together?
For an incremental migration, start small. Choose components that can be replaced in manageable steps and use a wrapper or compatibility layer to bridge the existing and replacement implementations. Maintain the required functionality while progressively updating outdated parts.
Clear boundaries make each step easier to integrate and verify. Decide what crosses the boundary, who owns the data and how failures are handled. The compatibility layer itself needs care, particularly where it connects components with different memory-management or security assumptions.
For a new architecture, the direction can be reversed: build the new foundation and integrate selected existing components around it. Both approaches allow reuse, but the architecture and objectives determine which should lead.
7. How will the improvements be demonstrated?
Define success in terms of the reasons for the upgrade. Security work needs evidence that the identified risks have been addressed. User-experience improvements need evaluation with users. Required compatibility needs checks against the behaviours and interfaces identified earlier.
Use the existing system as a reference where continuity matters, and use the new requirements where behaviour is deliberately changing. Plan deployment, recovery and support alongside implementation so that the organisation can operate the result with confidence.
A practical starting point
Begin with a concise statement of the problem, user feedback, the improvements required and the investment available. Compare a targeted fix, component replacement and architectural renewal against those needs. The outcome may be a modest repair or a substantial modernisation programme; either should have a reason proportionate to its cost and risk.