--- title: "Understand the Analytical Lifecycle" description: "The end-to-end flow that moves your work from raw data through built objects into production and back as monitoring data." --- Corridor follows a single end-to-end flow that moves your work from raw data through built objects, into production, and back again as monitoring data. The flowchart below walks through the full analytical lifecycle, from onboarding raw data, to building valuable features, models and policies in Pre-Production and to monitoring outcomes once they are live in production.
CORRIDOR · PRE-PRODUCTION CORRIDOR · POST-PRODUCTION YOUR STARTING POINT · OUTSIDE Your Data Source Lives in your stack today core banking · bureau · LOS data lake · TransUnion production scoring systems 01 Data Vault Register Data Onboard & catalog data tables · data columns metadata · lineage 02 Feature Eng. Build Features Transform data into variables DPDs · Ratios · Flags · WOE bins reusable shared feature library 03 Model Studio Develop Models Train · validate · champion vs challenger xgboost · sklearn · regression retrain · jupyter 04 Underwriting Compose Strategy & Rules Decision using data · features & models rules · approve · decline · cutoffs what-if scenario pricing · offer filtering 05 Simulate & Review Validate on historical data Model: AUC · KS · Decile lift Policy: Approval rate · bad rate waterfall 06 Approve & Export Sign-off · lock version · hand off Scorecards · JSON · fair lending change/audit log iterate ↻ RUNS IN PRODUCTION Your Production Env Outside Corridor LOS · Scoring engine · APIs Real-time + batch decisions e.g. MeridianLink · Fiserv Origence · Acxiom 07 Data Vault Bring Data Back In Close the feedback loop Decisions · Scores · Outcomes Dependent Variables 08 Monitor Performance Track model & policy health Model: PSI · CSI · KS · Drift Policy: Approvals · DPD · bad rate deploy refine · re-train model or policy
## What happens at each step
Step Module What you do
Pre-Production
01Register Data Data Vault Onboard source data tables and expose columns or aggregations as governed Data Elements with metadata and lineage.
02Build Features Feature Engineering Turn raw data into reusable shared variables (DPDs, ratios, flags, WOE bins) that any model or policy can pull from.
03Develop Models Model Studio Train models natively (xgboost, sklearn, regression, Jupyter) or import existing artifacts. Validate, retrain, and run champion vs challenger comparisons.
04Compose Strategy & Rules Underwriting Build a decision workflow on top of registered data, features, and models: approve/decline rules, cutoffs, pricing, offer filtering, and what-if scenarios.
05Simulate & Review All modules Run any object on historical data. Models surface AUC, KS, decile lift; policies surface approval rate, bad rate, and waterfall reports.
06Approve & Export All modules Send the locked artifact through a formal approval workflow with full change/audit log and fair-lending review, then export (scorecard, JSON) for deployment.
Post-Production
07Bring Data Back In Data Vault Pull decisions, scores, outcomes, and dependent variables from your production stack back into Corridor to close the feedback loop.
08Monitor Performance All modules Track model health (PSI, CSI, KS, drift) and policy outcomes (approvals, DPD, bad rate). When something drifts, refine or retrain and the loop starts again.
--- Next: [Start with a Model →](/getting-started/start-with-model/)