Predictions

Predictions is an advanced analytics feature powered by Loomi AI. It uses machine learning to forecast customer behavior—helping you identify the right audience for each campaign and maximize conversion rates.

How predictions work

Predictions combines historical and real-time customer data to evaluate each customer's likelihood of taking a specific action—such as making a purchase. You can use these propensity scores to segment your audience and target campaigns more precisely.

Prerequisites

For predictions to be accurate, you need enough historical customer data relevant to your chosen target. The more data you have, the more reliable the results.

Term	Definition
Customer	A customer profile in Bloomreach.
Data set	A collection of data points, also called samples.
Eligible customer	A customer who satisfies the conditions to be included in the model calculation.
Feature	An input variable the prediction model uses to make a prediction. For example, a customer's country or their number of purchases in the last year.
Feature window	The time period from which events are used to compute features. Expanding this window increases the information available to the model.
Model quality	A set of metrics estimating how accurately a model performs on new data—data that wasn't available during training. Common metrics include precision and recall (quantitative) and usability testing (qualitative).
Prediction model / decision tree	A model that uses statistics to predict outcomes.
Result / probability	The value returned by the model representing the predicted attribute.
Target	The variable the model is trying to predict — for example, a purchase.
Target window	The time period for which the model predicts the target — for example, the next 30 days.
Test data set	The data set used to evaluate prediction performance.
Training data set	The data set used to train the model.