Assessing industrial process innovation: development and application of a multidimensional composite model
Carlos César Ribeiro Santos, Leonardo Sanches de Carvalho, Lucas de Freitas GomesPurpose
This study aims to develop and apply the multidimensional industrial process innovation composite model (MIPICM) as a reproducible formative composite and examines what can be validly inferred from public project metadata.
Design/methodology/approach
The CORDIS Industrial Process Innovation Database comprises official Horizon 2020 and Horizon Europe projects. From 56,591 parsed projects, 16,094 met transparent eligibility criteria and 10,000 were selected by deterministic SHA-256 ordering. Frozen dictionaries transform auditable metadata into seven documentary-evidence dimensions. Evaluation includes a blinded 250-record human audit, missingness diagnostics, PCA, VIFs, 10,000-draw Monte Carlo weight sensitivity and cluster robustness.
Findings
Two independent evaluators achieved 98.76% agreement (κ = 0.963). Against exact human consensus, dictionary precision was 71.67%, recall 33.28% and F1 45.45%, indicating two independent evaluators achieved 98.76% agreement (κ = 0.963). Against exact human consensus, dictionary precision was 71.67%, recall 33.28% and F1 45.45%, indicating conservative retrieval. Weight sensitivity remained high (median rank correlation = 0.984; within-one-quintile agreement = 99.99%). D3 missingness was program- and time-structured, and clustering remained weakly separated and algorithm-dependent.
Research limitations/implications
Scores measure the intensity of documentary evidence, not realized productivity, overall equipment effectiveness, lead-time or carbon outcomes. External outcome validation and longitudinal evidence remain necessary.
Practical implications
MIPICM can support transparent screening and portfolio comparison when users retain links to the underlying records and avoid treating the index as a performance league table.
Originality/value
This study contributes an auditable, missingness-aware formative composite and demonstrates a conservative evaluation strategy for large public innovation-project databases.