Marketing & Media Mix Model - Glossary of Terms

This article is intended to serve as a glossary of terms that appear in the Marketing & Media Mix Model Dashboard.

Optimization Summary

  • Number of Scenarios: The number of budget scenarios for which optimization is performed. Each scenario corresponds to a fraction of the total baseline budget.
  • Marketing’s Contribution: The contribution of all non-marketing sources toward the target event. Check out our article Marketing & Media Mix Model - Marketing’s Contribution to get a more in-depth look at how we determine marketing’s contribution.
  • Baseline Spend: The sum of spend in all channels during the period defined by the baseline dates. This is the total budget for the baseline period and is used to determine the budget for each optimization scenario.
  • Baseline ___: The number of target actions (clicks, form fills, new customers, etc.) that occurred during the baseline period.

Validation Summary

This section starts with a few general definitions. These definitions are combined to address terms appearing in the dashboard.

General Definitions

  • XGB: Short for XGBoost - an open-source gradient boosting library. These algorithms learn underlying patterns from data by constructing a sequence of decision trees (think games of 20 questions). This is the algorithm used to build our Marketing and Media Mix Model.
  • Linear: A linear regression model also known as ordinary-least squares. A well-understood, interpretable model for regression that boils down to fitting a line to the data. This model is used as a baseline against which we compare our Marketing and Media Mix Model.
  • r2: R-squared or the coefficient of determination. This is a metric that measures how much of the variability in the data is explained by the model. The closer the metric is to 1, the more of the data variability is explained by the model and therefore the better the model. An R-squared of exactly zero corresponds to a model that predicts the average value of the target variable no matter the inputs. The R-squared can become negative which means the model is even worse than predicting the average value.
  • MSE: Mean-squared error. This metric corresponds to taking the difference between the predicted value and the actual value and squaring it for every date then dividing by the number of dates. The smaller the mean-squared error, the better.

Dashboard Definitions

  • XGB r2: The R-squared metric evaluated using our Marketing and Media Mix Model based on the XGBoost algorithm.
  • XGB MSE: The mean-squared error metric evaluated using our Marketing and Media Mix Model based on the XGBoost algorithm.
  • Linear r2: The R-squared metric evaluated using the Linear model. This is a reference to compare the XGB r2 value against.
  • Linear MSE: The mean-squared error metric evaluated using the Linear model. This is a reference to compare the XGB MSE value against.
  • Intercept: 1 - XGB r2. This is known as the unexplained variance and can be interpreted as the contribution to the target variable from factors besides marketing channels and seasonality.
  • Avg. Conversion Value: The average value of the target variable in your historical data.
  • Feature Importance: The importance our Marketing & Media Mix Model puts on each feature when predicting an outcome. The importance does not imply a positive relationship - here meaning higher importance leads to higher target values. It is just as likely that a feature can be important in decreasing target values. The importance simply quantifies which features the model pays the most attention to.
Was this article helpful?
0 out of 0 found this helpful
Have more questions? Submit a request

Comments

0 comments

Please sign in to leave a comment.