This tutorial uses deterministic synthetic data so the pipeline can be studied without NHS downloads.
python automation/pipeline_cli.py demo --months 3On Windows, RUN_DEMO.cmd runs the same command. Inspect:
work/demo_3_months/inputs/for contract-shaped source files;work/demo_3_months/pipeline.sqlitefor intermediate tables;work/demo_3_months/outputs/for analytical tables; andpipeline_validation_results.csvfor the gates.
Choose one practice and month. Compare it across OCS, GPAD, CBT, the union spine and a matched output. Notice that:
- source-led views keep the selected source even when another source is absent;
- the union spine preserves every reported source key;
- matched views contain only common eligible keys; and
- missing reporting is not changed to zero.
The analytical design index explains why each population differs.
The demo runs no clustering and its values are not NHS observations.