"Robustness testing is a frequently debated topic, particularly regarding how and when it should be performed. Guidelines are often interpreted to mean that robustness testing must be part of method validation, typically involving deliberate changes to method parameters to evaluate performance under normal usage conditions. In practice, this is commonly performed during the validation phase, although the timing and scope are dependent on the development phase and intended use of the assay.

When employing multivariate designs and total error approaches during method development, aspects of robustness are often already addressed. Furthermore, risk assessments—such as FMEA—can be used to identify parameters with the highest risk of assay failure. Designing experiments accordingly (e.g., using Analytical Procedure Control Strategies, or APCS) helps to build robustness into the method from the outset.

This raises the question: if robustness is already assessed through methodical risk-based development, is additional robustness testing during validation still necessary? In several cases, the combination of risk assessments, multivariate studies, and total error approaches has proven sufficient. Health authorities have accepted such evidence as adequate to demonstrate robustness, especially when well-documented and phase-appropriate.

In early development stages, the question of when to perform robustness testing depends on the specific needs and intended application of the assay. In a more risk-based, data-driven approach, the robustness section of the validation package is effectively informed by prior development activities and tailored to the phase-specific context."

Learn more from Dr. Lars Geurink by viewing his presentation "Analytical Procedure Control Strategy (ACPS)" during the Expert Forum: Overcoming key challenges in understanding and implementing guideline ICH Q14 for Analytical Procedure Development

Guidelines ICH Q2 and ICH Q14 specify that robustness testing is part of method development, although many companies make it part of the method validation plan. Ultimately, what is of interest for the company, is a good understanding and control of the analytical method. This knowledge and understanding comes from understanding the technological principles, the molecule, and from method development. 

My vote would be to perform robustness testing before method validation, indeed as part of method development and to an extend that you need to further your knowledge and experience of the method for the specific molecule. If you want to have it as part of the validation, ICH Q2(R2) does state that data from development can be lifted into validation ("Suitable data derived from development studies (see ICH Q14) can be used as part of validation data."). This would assume a proper quality level of your development data.