Pick a focus area, a statistical approach, and how the heavy computation should run. These choices scope the AI agents and models available downstream.
Optional, but recommended — without it, model choice falls back to a data-shape heuristic and narration stays generic.
3 — Statistical Model
Bayesian
Priors + uncertainty intervals — the MMM standard.
Random Forest
Non-linear, robust to messy/interacting features.
Decision Tree
Fast, interpretable, good baseline.
Gradient Boosting
Higher accuracy on structured/tabular data.
ARIMA / Prophet
Classical time-series forecasting.
Regularized Regression
Ridge/Lasso — simple, stable, explainable.
4 — Compute Backend
Heavy statistical work (Bayesian sampling, boosting, Monte Carlo, K-Means) runs faster with a real Python backend. Cloud (Colab) is free and usually faster than a local machine; Local keeps data fully on your own machine; In-Browser needs no setup at all. This choice applies to Statistical Analysis, Excel Model Builder, and Forecast Builder alike.
Checking backend…
Cloud (Colab): open stratos_advance_colab.ipynb in Google Colab, run all cells, paste the printed URL above — free, and usually faster than a local machine. Data is sent to that Colab session for the duration of the run, same as any cloud compute. Local: pip install fastapi uvicorn pandas numpy scikit-learn statsmodels scipy python-multipart then python stratos_advance_server.py — keeps data entirely on this machine, the right choice if a client's data can't leave their system.
5 — Run Analysis
Upload a CSV with a date column, a target column (sales/revenue), and one column per channel/promo.
☁ Share to Engine
Results
AI Narrative
Generating narrative…
Pharma Forecasting & Modeling
Bespoke pharma commercial models — moved here from Engine's Excel Model Builder and Forecast Builder. Same data pipeline as the tab above.
1 — Select Analysis Model
2 — Data Source
Upload a CSV/XLSX, or pull one already shared from Consulting/Engine.
☁ Share to Engine
3 — Workbook Details
All seven models run real math and export live-formula Excel sheets. K-Means, Monte Carlo and Competitive share their algorithms with Engine; Heatmap, Waterfall and Portfolio ROI are Advance-only.
1 — Forecast Mode
2 — Technique
3 — Reference Data (optional)
Not required — the technique below runs on the inputs you enter manually. Upload/import here only if you want a real dataset on hand to sanity-check those inputs against, or to share it onward.
☁ Share to Engine
4 — Inputs
Pull market data (optional)
Government payers only, gross of rebates (a floor, not total sales). Prevalence is model-based, US adults.
Runs on whichever backend is selected in Compute Backend (Statistical Analysis tab) — same JS math either way, Python/Colab just computes it faster.
Advance settings
Choose which model powers auto-model explanations and result narration for everyone using this app. Keys are stored centrally in Supabase, never in the browser.
Provider
Clients
Members who can sign in to this StratOS Advance workspace.
Loading members…
Save report for a member
They will see it under “My Saved Reports” and can download PDF, PowerPoint or Word.