Tipping Point¶
Tipping Point is a lightweight, high-performance marketing intelligence module that uses machine learning and calculus to determine the exact inflection points of a media response curve.
Inspired by the Google Meridian methodology, Tipping Point helps growth marketers make optimal, data-driven budget allocation and scaling decisions.
Key Concepts¶
The Hill Function A S-shaped (or C-shaped) media response curve that maps media spend to incremental return.
\[Return = \frac{\beta \cdot Spend^\alpha}{K^\alpha + Spend^\alpha}\]First Derivative (Marginal ROAS) Determines the marginal efficiency of the next dollar spent. Peak marginal efficiency occurs at the Inflection Point.
Tipping Points - Peak Efficiency Point (Daily): The inflection point ($f’’(x) = 0$) where acquisition cost is minimized. - Stop Scaling Point (Daily/Annualized): The point of diminishing returns where Marginal ROAS falls below your profitability threshold (default 1.0).
Geometric Adstock (Prior Decay Carryover) Accounts for lagged effects of prior spend on upcoming returns following a geometric decay model:
\[S_{t\_adstocked} = S_t + \theta \cdot S_{t-1\_adstocked}\]Tipping Point supports 4 adstock modes during training: - No Adstock: Assumes immediate return. - Free Adstock Fit: Automatically learns decay parameter $theta in (0, 1)$ from historical data. - Bounded Adstock Fit: Constrains the learned half-life decay to a specified day interval $[x, y]$. - Fixed Adstock: Directly enforces a user-defined decay half-life in days.
Portfolio Optimization (Cross-Channel Scenario Planning) Tipping Point scales from single-channel analysis to a full scenario planning engine. The PortfolioAllocator class takes multiple fitted MarketingReturnCurve models and utilizes scipy optimization (SLSQP algorithm) to find the exact budget allocation that maximizes total incremental return for a given total budget constraint, ensuring marginal ROAS is balanced across all valid channels.
Interactive Dashboard¶
Tipping Point includes a fully interactive Streamlit dashboard allowing web-based exploration, separated into two powerful stages:
Stage 1: Channel Configuration - Dynamic Stacking: Users can upload custom CSV data or input manual parameters to fit and stack multiple independent channels. - Value Denomination: Optional conversion value multipliers turn raw conversions (leads, installs) into revenue-denominated curves before fitting. - Deep Dive Analysis: Provides a channel-by-channel view of the Plotly saturation curve, marginal efficiency metrics, and an Adstock Carryover timeline displaying raw vs. accumulated spend.
Stage 2: Portfolio Optimization - Scenario Planning: Input a total portfolio budget and optionally set hard constraints (min/max limits) on specific channels. - Optimal Cross-Channel Allocation: Instantly calculates the most efficient distribution of funds across your configured channels. - Visual Benchmarking: Overlays all configured saturation curves on a single Plotly axis, cleanly marking the “setpoint” for each channel (solid line for funded spend, dashed line for untapped potential). - Scale Mix: A beautiful stacked area plot showing how your optimal channel mix expands, bottlenecks, and shifts weighting as your total investment ceiling increases.