TrueLift: Marketing Incrementality Lab
An interactive marketing measurement simulator that separates attributed conversions from genuine causal impact
TrueLift explores a question platform dashboards cannot answer alone: did advertising create the conversion, or simply claim credit for demand that already...
Synthetic, Reproducible Simulator: All customers, spending, campaigns, and outcomes are synthetic. The simulation is reproducible by seed, works without an API key, and uses optional AI only to improve explanations—not to generate or alter performance metrics. Forecastly is a fictional product; the demo is educational and not affiliated with any real company.
Why I Built It: Platform dashboards reward attributed conversions, but attribution is not causality. Marketing teams need a safe place to see how reported ROAS, last-click credit, holdout lift, and incremental LTV disagree—and what that means for budget, experimentation, and executive storytelling. TrueLift turns that measurement problem into an interactive lab rather than a static case study.
The Problem: Reported performance can look excellent while incremental impact is weak: ads often claim credit for demand that would have converted anyway. Last-click undervalues upper-funnel assist, platform claim factors inflate Paid Search and Meta, and finance cares about incremental CAC and LTV:CAC hurdles that attribution decks rarely surface together.
The Solution: TrueLift walks users through six connected views: Executive Overview (reported vs incremental KPIs), Audience (six personas), Attribution Lab (platform / last-click / experimental lift / incremental LTV), Experiment Builder (simulated holdouts with confidence intervals and power), Budget (channel reallocation under diminishing returns and marginal iROAS), and Measurement Council plus a printable executive Brief.
Default Scenario Insight: The default Forecastly scenario reveals how strong platform performance does not necessarily mean marketing created equivalent business value. Recommendations respond to profitability thresholds, marginal returns, recorded experimental evidence, and uncertainty rather than treating modeled outcomes as proven facts.
Measurement Council: A four-perspective Measurement Council brings together Attribution, Experimentation, Growth, and Finance viewpoints before synthesizing an executive-ready recommendation. Templates use the current budgets and recorded experiments; optional OpenAI (BYOK) only polishes narrative language, with Zod validation and template fallback if AI fails.
Architecture & Tech Stack: Built with Next.js App Router, TypeScript, Tailwind CSS, shadcn-style primitives, Recharts, and Vitest coverage on simulation formulas. A deterministic seeded engine in the simulation library generates coupled potential outcomes (with ads vs without), applies diminishing-returns spend curves, and encodes platform claim inflation, creator LTV multipliers, and promo retention haircuts. Scenario state is shareable via query params.
What I Learned: Incrementality is a decision system, not a single metric. Teams need persona-aware experimentation, hurdle-based budget logic, and honest uncertainty communication. The future of measurement tooling is operator-facing simulation that makes the gap between claimed credit and causal impact visible—and actionable—before real dollars move.
Educational simulator on deterministic synthetic data. Runs without an API key; optional OpenAI key improves Council explanations only and never changes performance metrics.