At a glance
- 1 Business understanding: targets and thresholds in config.yaml
- 2 Data understanding: quality score out of 100
- 3 Preparation: column mapping for real exports
- 4 Modelling: forecasts, exceptions, segments, reorder, roster
- 5 Evaluation: backtests, stability checks, R cross-check
- 6 Deployment: this site, Excel, SQL, Power BI, Tableau, WhatsApp, run report
How it works
- The quality score starts at 100 and loses points per issue (e.g. 0.5 per duplicate row, capped).
- The run report stores the SHA-256 of every input file, the config hash, and PASS/WARN/FAIL checks (freshness, beats the baseline, recalls blocked, P&L reconciles, cluster stability).
How to read it
- WARN means look before trusting a specific output. FAIL means don't act on that part.
Limits & safeguards
- The demo data is synthetic, generated to behave like a small Auckland convenience store.
Ask Shelly
How is the data quality?
Type or say these in the chat on the main page.