Simba documentation
SIMBA — Bayesian Marketing Mix Modeling (MMM) Platform
Simba is a no-code Bayesian Marketing Mix Modeling platform that measures media effectiveness, optimizes budgets, and forecasts marketing ROI. Replace spreadsheets, fragmented models, and black-box vendors with one transparent, enterprise-ready platform.
Built on the open-source PyMC framework by PyMC Labs, the team behind PyMC-Marketing, Simba combines the rigor of Bayesian statistics with an intuitive no-code interface — giving marketing teams enterprise-grade marketing mix modeling without writing a single line of code.
What is Marketing Mix Modeling?
Marketing Mix Modeling (MMM) is a statistical technique that measures the impact of marketing activities on business outcomes like revenue and conversions. Unlike last-click attribution or multi-touch attribution (MTA), MMM uses aggregate data to isolate the incremental contribution of each media channel — accounting for diminishing returns, carryover effects, seasonality, and external factors.
Simba makes MMM accessible to marketing teams who need rigorous measurement without hiring a data science team. See What is Marketing Mix Modeling? for a full explanation.
Why Choose Simba for Marketing Mix Modeling?
| Challenge | How Simba Solves It |
|---|---|
| Black-box MMM vendors deliver “trust me” results | Fully transparent — inspect every prior, parameter, and assumption |
| Custom MMM models take months to build and maintain | No-code model configuration with smart defaults — first model in 15 minutes |
| Fragmented tools for measurement, planning, and optimization | End-to-end platform: validate data, measure impact, forecast scenarios, optimize budgets |
| One-size-fits-all models ignore domain expertise | Bayesian priors let you encode business knowledge directly into the model |
| Siloed models across brands and markets | Portfolio modeling for cross-brand and cross-market consistency |
Key Features
Shared Studies and MCP Integration
Connect compatible AI clients to the same studies, recipe revisions, model runs and quality evidence used by analysts. The generated tool reference lists every tool of the current release, with capability discovery for the connected backend and analyst acceptance in the frontend. See the Simba MCP integration guide for setup, retry safety and connection refresh instructions.
Media Measurement & Attribution
Measure the true incremental impact of every marketing channel using Bayesian causal attribution. Integrate lift test results as likelihood observations to calibrate and validate your model. See Incremental Measurement.
Budget Optimization
Risk-adjusted budget allocation that accounts for saturation (diminishing returns), adstock (carryover effects), and uncertainty. Optimize across channels with configurable risk tolerance. See Budget Optimization.
Scenario Planning & Forecasting
Test budget scenarios before spending. Single-scenario prediction with uncertainty bands, what-if analysis, and carryover-aware forecasting. See Scenario Planning.
No-Code Model Configuration
Configure Bayesian priors, saturation curves, and adstock decay through an intuitive UI. Smart defaults auto-generate starting points based on your data and industry benchmarks. See Model Configuration.
Automated Data Validation
An AI-powered Data Validator checks your data across 10 validation categories before modeling — detecting anomalies, missing values, multicollinearity, and data quality issues. See Data Validator.
Portfolio & Multi-Brand Modeling
Cross-brand and cross-client modeling for agencies and multi-brand organizations. Consistent methodology, comparable KPIs, centralized management. See Portfolio Modeling.
Long-Term Effects (Bayesian VAR)
Measure long-term brand effects using Bayesian Vector Autoregression — impulse response functions, forecast error variance decomposition, and long-run equilibrium effects. See Long-Term Effects.
How Simba Works
Simba provides a complete workflow for marketing mix modeling:
1. Validate — The Data Validator automatically audits your data for quality issues before modeling.
2. Configure — Set up your model using no-code configuration with smart defaults or custom Bayesian priors.
3. Measure — Run the model and get incremental measurement of every channel’s contribution to revenue.
4. Optimize — Use budget optimization and scenario planning to allocate spend for maximum ROI.
The Bayesian Advantage
Simba uses Bayesian Marketing Mix Modeling rather than frequentist regression. This matters because:
- Uncertainty quantification — supported outputs come with intervals, including a 94% HDI (Highest Density Interval) on model coefficients and on each channel’s marginal ROI, so you know how confident to be before acting on a result
- Prior knowledge — encode domain expertise (e.g., “TV has longer carryover than paid search”) directly into the model
- Lift test calibration — integrate experimental results (lift tests, geo tests) as likelihood observations to validate and improve model accuracy
- Small data friendly — Bayesian models produce reliable estimates even with limited historical data
- Fully transparent — built on open-source PyMC, so every model component is inspectable and auditable
| Learn more: Bayesian Modeling Explained | Priors & Distributions |
Documentation
Getting Started
- What is Simba? — Product overview and positioning
- Quick Start Guide — Build your first marketing mix model in 15 minutes
- Account Setup — Registration, plans, and project configuration
- Platform Overview — UI walkthrough and navigation
Core Concepts
- Marketing Mix Modeling — What MMM is and why it matters
- Bayesian Modeling — The Bayesian approach to media measurement
- Incrementality — Causal attribution and incremental measurement
- Saturation Curves — Diminishing returns and response curves
- Adstock Effects — Carryover, memory decay, and lagged impact
- Priors & Distributions — Configuring Bayesian priors
- Seasonality — Seasonal patterns and trend modeling
Platform Guide
- Data Validator — Automated data validation and quality scoring
- Model Configuration — Configuring priors, saturation curves, and adstock decay
- Smart Defaults — Auto-generated model starting points
- Incremental Measurement — Channel attribution and contribution analysis
- Budget Optimization — Risk-adjusted budget allocation
- Scenario Planning — Forecasting and what-if analysis
- Long-Term Effects — Bayesian VAR for brand equity modeling
Data
- Data Requirements — What data you need and supported formats
- Data Preparation — Cleaning and formatting best practices
- Data Validation — How the Data Validator audits your data
- Supported Channels — TV, digital, social, OOH, and more
Use Cases
- Brand Marketers — For in-house marketing teams
- Agencies — Multi-client management and portfolio modeling
- Portfolio Modeling — Cross-brand and cross-market analysis
- Retail & E-commerce — Online and omnichannel retail
Security & Compliance
- Security Overview — AES-256 encryption, TLS 1.2+, Cyber Essentials certified, GDPR compliant
Resources
- Glossary — Marketing mix modeling and Bayesian statistics terminology
- PyMC-Marketing & Simba — How the open-source project and platform relate
- Further Reading — Papers, articles, and external resources
- FAQ — Frequently asked questions
- Pricing & Plans — See getsimba.ai for current plans
Simba vs. Other MMM Solutions
| Capability | Simba | Google Meridian | Meta Robyn | Custom In-House |
|---|---|---|---|---|
| No-code UI | Yes | No (Python) | No (R) | No |
| Bayesian framework | Yes (PyMC) | Yes (lightweight Bayesian) | Ridge regression | Varies |
| Uncertainty quantification | Intervals on supported outputs | Limited | No | Varies |
| Budget optimization | Built-in, risk-adjusted | Separate | Basic | Build your own |
| Lift test integration | Yes (likelihood observations) | Yes | Yes (calibration) | Build your own |
| Portfolio / multi-brand | Built-in | No | No | Build your own |
| Long-term effects (VAR) | Built-in (Bayesian VAR) | No | No | Build your own |
| Enterprise security | Cyber Essentials, GDPR | Google Cloud | Self-hosted | Self-managed |
| Time to first model | 15 minutes | Days–weeks | Days–weeks | Months |
For a detailed comparison against other tools, talk to us.
Getting Help
- GitHub Issues — Bug reports, feature requests, and support questions
- Email: info@pymc-labs.com
- Website: getsimba.ai
- Start free: demo.simba-mmm.com/users/signup
- Book a demo: Schedule a call
Built on PyMC
Simba’s models are written in PyMC, the open-source probabilistic programming framework, by PyMC Labs, the team behind PyMC-Marketing. The engine is Simba’s own; it does not import PyMC-Marketing. This means:
- Full transparency — the probabilistic models driving your ROI are inspectable and auditable
- Scientific rigor — built on decades of Bayesian statistics research
- No vendor lock-in — your modeling logic is built on open-source foundations
- Community-driven — benefit from continuous improvements by the PyMC community
License
Copyright 2026 1749 Ltd. All rights reserved.
This repository and its contents are proprietary. See LICENSE for details.
| Privacy Policy | Terms of Service |
Simba — Bayesian Marketing Mix Modeling platform. Built on PyMC by PyMC Labs, the team behind PyMC-Marketing.