Service
A forecasting, risk or ranking model delivered as an application your team uses, with its assumptions stated, its validation shown and scenario controls in the interface.
Most models never leave the notebook they were built in. The analysis is sound, but the people who make the decision cannot change an assumption and see what moves.
When a model does reach them, it often arrives as a single number without the conditions it depends on, so it is trusted too much or not at all.
A decision model is useful when its assumptions are visible, its accuracy is measured on data it did not see, and the person deciding can run their own scenarios.
Sliders and inputs for the assumptions that matter, with results that update in place.
Accuracy reported on data the model did not train on, with the method stated.
Where causality is in question, methods with different assumptions are compared; disagreement is reported, not hidden.
Each input's source, period and treatment documented next to the result.
Data refreshed on a schedule that matches how fast it decays, with the date shown.
What the model does not cover is part of the interface, not a footnote.
Decision Models
LiveRanks all 21 PJM zones by grid vulnerability as data center load grows, with three methods that check each other.
Decision Models
Wakes on demandCompares stocks and ETFs on risk-adjusted terms in one view, with the statistics computed by a separately tested API.
The decision comes first: what is chosen, by whom, how often, and what error costs. The model is sized to that.
Out-of-sample tests, and where relevant event studies or fixed effects that separate cause from coincidence.
Scores keep their meaning across scenarios, so one scenario can be compared with another.
An application with an API behind it, so the statistics can be tested apart from the interface.
Operate for one model and its application; Scale when several models share data or feed each other.
Useful add-ons: Dashboard; Data pipeline.
Siting, pricing, allocation, risk limits, forecasts that drive staffing or purchasing: decisions made repeatedly from data that changes.
No. Decision models in finance describe risk and return under stated assumptions. The decision and its suitability stay with your firm.
On held-out data, with the metric chosen for the decision, and with the conditions under which the model should not be used.
You do. Code, data pipelines and documentation are yours at the end of the engagement or the plan.
Describe the process and the data. A reply comes within one business day, from the person who would do the work.