Decision to improve
How should limited resources be allocated when outcomes depend on many interacting factors and the final plan must obey real constraints?
What shapes the outcome
Teams often need to decide where limited time, money, or capacity will have the greatest effect. The challenge is that outcomes depend on interacting factors, only some can be changed, past allocation may not reveal true opportunity, and every recommendation must fit real policy and capacity limits.
Constraints and uncertainty
- Predicting an outcome does not automatically show which action will improve it.
- The effect of one action can change across people, conditions, and combinations of actions.
- Simulated options become unreliable when they stray too far from situations represented in the data.
- The recommended allocation must be feasible, explainable, and measurable after it is put into practice.
How the system works
- Represented the desired outcome, what the team could change, what it needed to account for, and the constraints on any plan.
- Estimated how outcomes responded to different actions, using temporal validation and diagnostics for interactions, data support, and stability.
- Simulated realistic scenarios before resources were committed and kept proposed changes within the range supported by the data.
- Optimized the allocation under real constraints, then paired the plan with a way to measure results and improve the next decision.
What this system enables
The result is a reviewable action plan rather than a passive prediction. TreeMMM publicly demonstrates the pattern through nonlinear response modeling, interaction discovery, scenario curves, and bounded budget reallocation on customer-level panel data, with a feature-parity R companion.
What I built
I designed and built the end-to-end decision and optimization workflow, including modeling, interpretation, simulation, allocation, and measurement, drawing on applied work in targeting, forecasting, causal measurement, and resource optimization.