MIM Overview | Glossary of Terms
Baseline Window: A window of past dates used to show historic performances from your OneView. The windows are defined to have the same number of days as your Planning Window.
Cluster Feature: Should recommendations be clustered into groups of similar recommendations? This reduces thousands of recommendations to a more manageable number. Clustering is done by default.
Diminishing Return on Ad Spend: At a given level of spend, how saturated is the market or, equivalently, how greatly does your return diminish.
Marginal Return on Ad Spend: Historically, the return in a stage of your pipeline for every $1 spent
Maximize Target Prediction: Recommendations provide a larger value of your target than the baseline, but are less efficient than your baseline.
Number of Clusters: If clustering is enabled, how many final groups would you like?
Planning Window: A window of future dates over which the model will perform optimization. Each window begins on the first day of the next month and extends for varying lengths of time.
Prediction Categories:
- Top Prediction: A recommendation that provides a larger value of your target for a smaller cost when compared to your baseline. Additionally, the recommendation with the best Cost Per of all recommendations is counted as a Top Prediction. For more information, check out our Prediction Categories Explained article.
- Great Prediction: Any recommendation that has a lower Cost Per than the baseline scenario.
- Not a Recommended Prediction: Recommendations that are inferior to other predictions and should not be used.
Prediction Performance: How well does your model perform when asked to predict things that have already occurred? The SMAPE metric (symmetric mean absolute percentage error) is used here. For more information, check out our Can I use my model if it's performing below benchmark?
ROAS: Total Sales Revenue / Total Marketing Ad Spend
Cost Per: For impressions, this is calculated as CPM i.e. (Spend / Total Impressions) * 1000