LIGHTWELL
LIGHTWELL
Causal Intelligence PlatformLive Data

Research Data

Nine production-deployed modules spanning the full marketing science stack — from real-time streaming attribution through Bayesian media mix modeling, geo-lift incrementality, and causal experimentation. Every system is publicly accessible with a live dashboard and, where available, a DOI-indexed whitepaper.

Platform Metrics

208Kevents/sec
Streaming throughput
Artemis engine
87msp99
Attribution latency
exactly-once semantics
98.2%
Signal integrity
privacy-safe ID resolution
R-hat 1.02
MMM convergence
94.2% model confidence
92.4%P90
Forecast intervals
64 H3 zones
35.4%variance ↓
CUPED reduction
82.4% power maintained
R² 0.92
Geo-matching fit
Mahalanobis matching
18.4Mevents
Validated audience
4 regions

Live Systems Index

9 modules · production-deployed · Vercel
Module 01LIVE

Artemis Streaming Engine

Incremental Markov-Shapley attribution · 208K events/sec · 87ms p99

Live DashboardReal-Time Streaming Attribution
Module 02LIVE

Probabilistic ID Resolution

Cross-device household entity resolution · GMM clustering · 98.2% signal integrity

Live DashboardProbabilistic Identity Resolution
Module 03LIVE

Media Mix Optimizer

Bayesian hierarchical MMM · Hill saturation · adstock · NUTS/HMC sampling

Live DashboardBayesian Media Mix Modeling
Module 04LIVE

Event ROI Command

Live broadcast-to-mobile attribution · mSPRT sequential testing · Welford spike detection

Live DashboardLive Event Attribution
Module 05LIVE

Calibration Test Suite

Causal calibration across the full modeling stack

Live DashboardThe Causal Calibration System
Module 06LIVE

Psychographic Intent Hub

Behavioral profiling and causal uplift modeling

Live DashboardBehavioral Profiling & Causal Uplift
Module 07LIVE

Urban Strategy Engine

Spatio-temporal demand forecasting · Prophet + ARIMA + LSTM · H3 hex indexing

Live Dashboard
Module 08LIVE

Decision Intel Platform

CUPED variance reduction · SPRT with O’Brien-Fleming boundaries · SRM detection

Live DashboardIncrementality Testing at Scale
Module 09LIVE

Synthetic Lift Studio

Synthetic Control + DiD · Mahalanobis geo-matching · placebo-based inference

Live DashboardThe MMM-Incrementality Bridge

Published Whitepapers

10 papers · DOI-indexed · Zenodo

Formal technical whitepapers with mathematical derivations, published with persistent DOIs through Zenodo. Papers a–d carry their original v1 DOIs; papers e–j resolve to the LaTeX-compiled v2 versions.

a

A First-Principles Hybrid Attribution Framework

10.5281/zenodo.18557680
b

Bayesian Media Mix Modeling: Axiomatic Budget Optimization

10.5281/zenodo.18599386
c

Behavioral Profiling and Causal Uplift

10.5281/zenodo.18599425
d

The Causal Calibration System

10.5281/zenodo.18599433
e

Probabilistic Identity Resolution

10.5281/zenodo.18860338
f

Live Event Attribution

10.5281/zenodo.18860339
g

Real-Time Streaming Attribution

10.5281/zenodo.18860342
h

Incrementality Testing at Scale

10.5281/zenodo.18860345
i

Marketing Data Connectors

10.5281/zenodo.18860349
j

The MMM-Incrementality Bridge

10.5281/zenodo.18860350

Technical Skills

Causal Inference

Markov chains · Shapley values · SCM · DiD · uplift modeling · Meta-Learners · DML · mSPRT

Bayesian Methods

NUTS/HMC sampling · hierarchical priors · Hill saturation · adstock decay · R-hat convergence

Forecasting

Prophet · ARIMA · LSTM · ensemble weighting · H3 geospatial indexing · uncertainty quantification

Experimentation

CUPED · sequential testing · O’Brien-Fleming · SRM detection · power simulation · placebo tests

Real-Time Systems

Stream processing · exactly-once semantics · Welford’s algorithm · <100ms latency pipelines

Frontend

Next.js 15 · React 18 · TypeScript · Tailwind CSS · Recharts · responsive dashboards

Backend

Python · FastAPI · NumPy · SciPy · pandas · Node.js

Deployment

Vercel · GitHub Actions · Docker · CI/CD

Experience

Founder & Lead Engineer — Causal Intelligence Platform

Jan 2026 – Present

Forsythe Publishing & Marketing

  • Designed and built a 9-module platform spanning the full marketing science stack: streaming attribution, identity resolution, Bayesian MMM, live event ROI, causal calibration, behavioral uplift, forecasting, experimentation, and geo-lift incrementality.
  • Artemis Streaming Engine: incremental Markov-Shapley attribution at 208K events/sec, 87ms p99, exactly-once semantics via checkpointing.
  • Media Mix Optimizer: Bayesian hierarchical model with Hill saturation and geometric/Weibull adstock; NUTS/HMC sampling, R-hat 1.02.
  • Produced 10 technical whitepapers with formal mathematical derivations. All systems publicly deployed on Vercel.

Founder & CEO

2024 – Present

Forsythepublishingandmarketing.agency

  • Built an AI voice receptionist service for healthcare, hospitality, and fitness verticals — LLM-powered conversation with structured intake workflows.
  • Developed lead generation pipeline and go-to-market strategy targeting medical practices. Bootstrap-funded, zero external capital.

Founder

2023 – Present

Forsythe Publishing & Marketing

  • Built and operated digital marketing and content publishing operations. Developed internal attribution and analytics tooling that became the foundation for the Causal Intelligence Platform.
MF

Michael Forsythe Robinson

Marketing Science Engineer · Causal Intelligence & Attribution Systems · Fort Worth, TX

github.com/Michaelrobins938LinkedInORCID 0009-0002-8487-759Xmichael.robins938@gmail.com
Author Page ↗

Explore the research behind the platform

Read the whitepapers, open the live dashboards, or review the scientific method driving every module.