Measuring Avoidable Delivery-System Friction in Enterprise Platform Engineering

Replication package for a longitudinal, embedded multiple-case study of avoidable delivery-system friction in two large regulated U.S. enterprises (420 teams; 3,407 anonymous survey responses; 2017–2025). The workbook computes the Enterprise Cognitive Load Index (ECLI), a prespecified formative composite of survey effort, support burden, onboarding friction, and feedback latency, from published aggregate inputs alone. It contains the case study protocol, data dictionary, raw-to-results lineage, all 32 ECLI values, weighting and normalisation sensitivity analyses, exploratory statistics, per-question analysis sheets, and a Python verification script reproducing every published value. No individual survey responses, participant identifiers, or raw enterprise telemetry are included. Contents ECLI_EMSE_Replication_Workbook_v2.xlsx — 32 sheets; no macros; opens in Excel, Excel for the web, LibreOffice, and Google Sheets. verify_ecli.py with run instructions and expected outputs. Reproducing the results Place both files in one folder and run python verify_ecli.py (requires pandas, numpy, scipy, openpyxl). The script reads only the two aggregate input sheets and recomputes everything else independently. Published values reproduce to a maximum absolute deviation of 1.42 × 10⁻¹⁴. Interpretation boundaries T, O, and F components use organisation-specific normalisation spanning both measurement windows, so published scores are retrospective by construction and cross-organisation point comparisons are scale-dependent; rank-based results are not. Health bands are operational heuristics, not validated cut-points. Every inferential test is exploratory: capability areas are fixed nested strata, so no p-value supports population inference and none is causal. Arithmetic is fully auditable; source-record extraction is not. The package reports negative results and enumerates fourteen limitations that cannot be remedied from the retained evidence. Licence Data are licensed CC BY 4.0. The verification script (verify_ecli.py) is separately licensed under the MIT licence; full text is included in the archive.