What is the documented return on investment for research data infrastructure across studies?

A subsidiary question under Q-0004. The cost-side of the architectural argument needs an empirical anchor on the value side. Q-0014 asks what the existing literature actually measures.

The answer is C-0031: positive ROI in every documented study, ranging from 5:1 (UK Archaeology Data Service) and 7:1 (Australian NCRIS) at the conservative end to 18-88:1 (XSEDE) and 20-26:1 (EMBL-EBI) in the cluster, with the Protein Data Bank at 800:1 as the documented outlier. Apon et al. measured every $100K in research-computing salaries as associated with $14.3M increase in HERD expenditure.

C-0032 extends this with the open-data multiplier: when preserved data can be openly shared, returns compound further (Landsat 53→5,700 scenes/day after open access; HGP $14.5B → $965B; PDB facilitating 88-100% of recent FDA drug approvals; SARS-CoV-2 → $1.9T Pfizer-BioNTech vaccine value within months of public sharing). The EU estimated the cost of not having FAIR research data at €10.2B/year minimum.

The measurements underestimate the architectural recommendation's return because they are measured against regimes that already truncate at the grant cycle (Q-0042).