From composite indicators to DataOps backbones: a mathematical architecture for scalable SME innovation diagnosis

This paper extends the OECO\_Sy model for SME innovation propensity from a determinant-level composite indicator to a scalable data architecture for university-led open innovation ecosystems. The starting point is the submitted companion paper by Di Letizia and Grilli, which provides a previous mathematical formalization of three coefficients, namely the Preliminary Innovation Propensity Coefficient (PIPC), the Management Innovation Propensity Coefficient (MIPC), and the Composite Coefficient of SME Innovation Propensity (CCSIP). The new contribution is the definition of a DataOps-oriented backbone, denoted by OECO\_SyMDB, which makes the diagnostic process repeatable, traceable and extensible over time. We model the architecture as a family of source-indexed and time-indexed data objects, a versioned composition of transformations from landing to exploitation zones, and a deterministic Data Analysis Backbone that reproduces the OECO\_Sy coefficients from the questionnaire sub-view only. This distinction corrects a common methodological ambiguity: the multi-source architecture may contain questionnaires, balance sheets, reports, innovation projects, industrial property records and briefs, but PIPC, MIPC and CCSIP must be computed only from the validated questionnaire determinants. Additional sources generate second-level indicators, not altered baseline coefficients. Several propositions establish boundedness, monotonicity, reproducibility under a fixed methodological version, and horizontal extensibility under schema invariance. A pilot proof-of-concept on the ten anonymized TCOREC firms is used only to verify deterministic recalculation of the coefficients, not to claim end-to-end validation of the complete five-source pipeline. The result is a mathematically controlled architecture for transforming a single-firm diagnostic model into a longitudinal observatory of SME innovation propensity.