Models Overview
Models are a crucial tool in understanding the factors that drive your business's sales. By building statistical models of your factor data, you can gain insight into how each factor impacts your sales and create plans that maximize profitability.
Section Index
- Models Home Page
- Archive & Delete Models — An outline of the two-stage process for model management, explaining how to first temporarily archive models and then permanently delete them.
- Published Model — An overview of the "publish model" feature, explaining how it designates a specific model as the official source of truth, making it the default for all plans and easily switchable.
- Model Comparison
- Building a Model — An outline of the process for creating a new model, detailing the initial setup of its date range and factors, taking a data snapshot for quality control, and running the final simulation.
- Model Requirements — A summary of the specific data requirements for model creation, outlining the necessary inputs for Sales (Outcome), Marketing Factors (Tactics), and optional Environment factors.
- Model Timeline Best Practices
- Data Snapshots — An explanation of how model data snapshots are created from the Factors & Data library, covering the flexibility of point-in-time data, customization options, and the meaning of data warning alerts.
- Add Your Research — An overview of how to incorporate your own ROI and elasticity research into the model, covering the conversion process, the Bayesian updating method that combines it with your data, and special considerations for portfolio-level analysis.
- Model Holdout Testing — A guide to using the holdout test to validate model predictivity, detailing the testing process, how to assess results, and common quality control troubleshooting steps.
- Monte Carlo Simulation
- Building a Portfolio Model — Combining single-segment models with portfolio tactics to build a comprehensive portfolio model.
- Editing an Existing Model — An outline of the locked models feature, detailing which changes are permissible in a locked model versus those which require creating a new model.
- Adding New Models — An overview of the impetuses for creating a new model, and the considerations to make when building it.
- Model Results Brief & QC Walkthrough — An overview of the Model Results Brief, explaining how to interpret its key charts and statistics - such as historical prediction, MAPE, R-square, and factor estimates - to validate a model's performance.
- Retention
- Marketing Factor Statistics and Impact — A set of answers to common questions about interpreting marketing factor statistics, defining Mean Estimate and Standard Error, and recommending which metrics to use for evaluating a factor's overall impact.
- System Suggested Factors — An overview of the System Suggested Factors feature, which analyzes model errors to recommend new factors, allowing users to iteratively add them to improve model accuracy.
- Measuring Long-Term Effects
- Distinguishing Incremental Sales from Base