bhupesh-debug/station-forecast-reconciliation ? reverse-engineered prompt

Build me a Python project that helps a broadcast FP&A team make monthly forecasts line up across 17 stations and the enterprise total.

I want the app to generate synthetic station and enterprise data, run a few forecasting approaches like bottom up, top down, and a reconciliation method that tries to make everything consistent, then compare them in a simple backtest. It should also flag stations whose forecasts look off, and include a small Streamlit dashboard so an analyst can quickly see where the biggest mismatches are.

Please also add two extra pieces, one that estimates which CapEx projects are likely to overrun, and one that shows how much political election cycles and major sports events affect ad revenue. Keep the results reproducible, save charts and summary outputs, and include tests for coherence and no lookahead issues. Use current docs online if you need to, and make the project easy to run from the command line.