Quais são algumas armadilhas comuns de requisitos não funcionais e como evitá-las?

Using a generative artificial intelligence to help with rewording your technical documentation using the best practices mentioned above. Just use a little prompt engineering.

To clearly define and document non-functional requirements (NFRs), engage stakeholders to gather input and categorize NFRs (e.g., performance, security). Use clear, specific language and measurable criteria, such as "The system should handle 5000 concurrent users with a response time under 2 seconds." Provide context and rationale, use standardized templates, and include acceptance criteria. Regularly review and update the documentation, ensuring it is accessible and validated by stakeholders.

NFR's should be quantified in the same way that any other application requirement would be. The acceptance test criteria and test cases related to the NFR's need to be defined, otherwise there is no way to validate that the system meets the requirements at the end.

And overlooked! There are some requirements that are so obvious to those writing them down that it never occurs to them to write them down. Royal Navy submarine sonar array - "But we wanted it to be black!" meant a complete redesign.

To handle contradictory non-functional requirements (NFRs), identify contradictions early and engage stakeholders to discuss them. Prioritize requirements using techniques like MoSCoW and perform trade-off analysis to understand impacts. Negotiate compromises and document decisions with clear rationale. Continuously review NFRs in each iteration, prototype and test solutions, and seek expert advice if needed. Maintain flexibility to revisit decisions as the project evolves, ensuring balanced and well-informed outcomes.

For many things, especially IT, interactive performance is a critical element of the user experience, yet this is often not tested. Can you imagine driving a car where the steering had a lag in response? How many times have you heard someone in a contact centre say "I'm sorry, the system is really slow today"? Always ensure that interactive response performance requirements are clearly defined, realistic and thoroughly tested! This has avoided problems for me on a number of occasions.

Use a tool like Azure DevOps to document the nonfunctional requirements and have them ready to address as soon as possible when the sprint backlog asked. Why? This proactive approach minimizes technical debt.

Design your DevOps CI/CD pipelines to use tooling like Application Insights to measure system KPIs regularly and investigate anomalies. Many management interfaces allow you to set alerts or even scale automatically.

Build for accessibility. Take the time to learn more about WCAG and other accessibility initiatives. This is too often an afterthought and not built in from the beginning. With generative AI making an interface more accessible is easier than even.