Skip to content

PRICING_SCENARIOS.md

Latest commit

 

History

History
53 lines (39 loc) · 1.86 KB

PRICING_SCENARIOS.md

File metadata and controls

53 lines (39 loc) · 1.86 KB

Pricing Scenario Runs

The pricing scenario runner turns the /simulate/pricing contract into a persisted decision artifact. It evaluates a portfolio of incentive levels across income segments, writes the scenario table, and adds sensitivity rows for margin downside/upside and vulnerable-customer-share stress.

Run

uv run python -m src.pricing.scenario_runs --run-date 2025-06-30

With a custom incentive grid:

uv run python -m src.pricing.scenario_runs `
  --run-date 2025-06-30 `
  --incentives 0 2 4 6 8

Default output:

artifacts/pricing/scenario_runs/run_date=2025-06-30/pricing_scenario_run.json
artifacts/pricing/scenario_runs/run_date=2025-06-30/pricing_scenario_run.md
artifacts/pricing/scenario_runs/run_date=2025-06-30/pricing_scenarios.csv
artifacts/pricing/scenario_runs/run_date=2025-06-30/pricing_sensitivity.csv

Inputs

When neobank.duckdb exists and contains the dbt marts, the runner derives segment priors from:

  • main_marts.fct_activation for eligible customers, D7 activation, and vulnerable-customer share.
  • main_marts.fct_user_clv_proxy for expected monthly margin assumptions.
  • main_marts.mart_pricing_recommendations indirectly through the API scenario simulator when pricing-mart evidence is available.

If the DuckDB file is missing, the runner falls back to conservative synthetic segment priors so the public repo remains reproducible without generated data.

Decision Policy

  • ship: candidate for a controlled rollout, subject to monitoring.
  • iterate: candidate for a smaller experiment or cheaper incentive design.
  • hold: stop state until economics or guardrails improve.

Review the sensitivity CSV before choosing a scenario. A scenario that looks good only in the base case but fails under margin downside or vulnerable-customer stress should stay in experiment design, not rollout.