Introduction. The pharmaceutical industry is transitioning from quality-by-test (QbT) to quality-by-design (QbD) and quality-by-digital-design (QbDD) paradigms, necessitating built-in quality assurance throughout the product lifecycle [1]. High-fidelity mathematical models are critical for reliable process design, optimisation, and control [2]. However, robust model establishment is challenged by scarce experimental data and inadequate assessment of data information content, hindering parameter estimation and undermining predictive capability [3]. Crystallisation processes exemplify this difficulty, as complex mechanisms (nucleation, growth, agglomeration, polymorphism) yield high-dimensional population balance models with extensive parameter sets, which demands substantial experimental resources yet often yields information-deficient data [4,5].

Methods. We developed ESTANPy, a Python-based web application integrating global sensitivity analysis with sequential orthogonalisation to quantify data information content, diagnose non-estimable parameters, and guide information-rich experimental design.

Results. Applied to a 16-parameter paracetamol batch cooling crystallisation model, ESTANPy identified 10 estimable parameters from preliminary data; estimability-guided model-based design of experiment (MBDoE) subsequently yielded an optimally designed experiment, increasing this to 12 [4]. Compared with traditional full factorial design requiring a minimum of 16 runs, the proposed approach achieves equivalent estimability with 3 runs (more than 80% reduction in experimental burden), whilst markedly curtailing material consumption, energy usage, and environmental impacts.

Conclusions. ESTANPy enables efficient, targeted model development by identifying information-deficient parameters and guiding information-rich experimental design. This capability directly supports QbDD implementation in pharmaceutical processes, delivering substantial reductions in experimental burden without compromising model reliability.