Demand moves before you see it. Thagorus sees it first.
Weather reshapes what people buy, when they buy it, and how much they’ll pay — across thousands of categories, in ways no rule-based system can predict. Thagorus is a causal demand model that turns those invisible patterns into precise ad-spend recommendations, down to the dollar.
Modular demand intelligence, end to end.
From atmospheric conditions to causal inference to budget recommendations — every layer is inspectable, every number is auditable.
Rules see temperature. The model sees demand.
Same temperature, opposite outcomes. Wind, humidity, UV lag, rate of change, day-of-week interactions — the variables that actually drive purchasing behavior are invisible to threshold rules. Thagorus encodes 12 weather dimensions into a causal demand model that captures the nonlinear, counterintuitive relationships between atmosphere and commerce. The result: dramatically tighter predictions where it matters most — at the inflection points rules miss entirely.
Average forecast error with rule-based weather triggers
Thagorus causal model with multi-variable encoding
Simulated dataThis is what a Thagorus recommendation looks like.
Not a dashboard. Not a score. A complete evidence bundle: the weather signal that triggered it, the causal estimate behind it, the confidence interval around it, the conditions under which it breaks, and the backtest that validates it. Every recommendation is auditable. Every number is clickable.
Weather changes. Demand follows.
Four capabilities, one system. From live atmospheric conditions to category-level demand signals to auditable evidence bundles — everything connected, everything inspectable.
Performance marketing teams that run on data.
Designed to work with your existing stack — connect to Meta, Google, Amazon, and programmatic platforms via API.
The demand layer the industry is missing.
Today, Thagorus is in private beta. The causal identification methodology is designed. The multi-tenant pipeline is being built. The first design partners are contributing data.
What exists today
- Weather-shock identification using natural experiments across markets
- Hierarchical partial pooling across tenants and categories
- Evidence bundles with confidence intervals and break conditions
- Shadow mode deployment with human-in-the-loop by default
What we’re building
- Deep learning demand models with cross-attention weather encoding
- Time-series foundation model integration for zero-shot new category onboarding
- Reinforcement learning budget optimization with weather-contingent policies
- Federated learning for privacy-preserving cross-tenant intelligence
The science is real. The math is proven. The market timing is right — AI weather forecasting accuracy improved 10x in the last 18 months, and the tools to build this finally exist.
See a proof bundle with your category.
Tell us what you sell and where. We will show you what the model sees.
Or email nate@schmiedehaus.com directly.
Currently accepting design partners in outdoor, beverage, apparel, personal care, home improvement, and food delivery categories.