Description

This regression model provides the value of each customer in a given window of time in a brand’s customer base. This is calculated using the following metrics:

  • HISTORICAL_LTV: all spend prior to score date
  • SCORE: the predicted spend of a customer in the PREDICTION_WINDOW
  • TOTAL_LTV: the sum of HISTORICAL_LTV and SCORE

HISTORICAL_LTV is the past spend, SCORE is the predicted future spend, and TOTAL_LTV is the combination of both. All three can be used for audience creation and segmentation, depending on the goal.

Model Time Windows

  • PREDICTION_WINDOW: 365 days
  • HISTORICAL_WINDOW: 365 days

Predictable’s default value for the window of prediction (known as the PREDICTION_WINDOW) is 365 days. This means the model is predicting the customer’s spend over the next 365 days. If necessary, we are able to assign a custom PREDICTION_WINDOW that fits your business’s unique sales cycle. Each customer’s unique data is transformed into a custom feature set, with automated training and tuning to deliver the most performant model possible.

The HISTORICAL_WINDOW is how many days Predictable’s models look back for data. To be scored by the LTV model, a customer must have made a purchase, clicked an email, or triggered a web event within the HISTORICAL_WINDOW. The default value for LTV’s HISTORICAL_WINDOW is 365 days .

After training and tuning, the model is ready to score the active population.

Data Processed:

  • Transaction Data
  • Email Engagement Data
  • Web / Pixel Engagement Data

Output

Model Results

The model produces four different columns, all of which have different applications for your marketing efforts.

Returned Values:

  • HISTORICAL_LTV: all spend prior to score date
  • SCORE: the predicted spend of a customer in the PREDICTION_WINDOW
  • SCORE_SCALED: the score transformed into a ranked scale – 100 is the most valuable customers during the window of prediction, 1 is the least valauble
  • TOTAL_LTV: the sum of HISTORICAL_LTV and SCORE
  • CUSTOMER_ID: your unique customer identifiers
  • DATETIME_STAMP: unix timestamp of scoring run
  • MODEL_VERSION: version of platform that scored the run

Model Summary

Additionally, Predictable returns model summary statistics for you to assess how well the model fits your data.

Returned Values:

  • R2: a metric that indicates how much variability within the test data the model explains. This value is between zero and one – the higher, the better
  • RMSE: (Root Mean Squared Error) a metric that evaluates the average difference between the predicted spend and the actual spend in the test set. The lower the value, the less the average error the model
  • MAE: (Median Absolute Error) a metric that evaluates the median absolute difference between predicted spend and the actual spend on the test set. The lower the value, the less the average error the model
  • NUMBER_OF_FEATURES: number of features in the model train
  • TIMESTAMP: unix timestamp of training run
  • MODEL_VERSION: version of platform that trained the model

Feature Importance

Predictable also provides the relative importance of the features (inputs) of the model. The higher the score, the more important the feature was to the model. However, it is extremely important to note that the importance value does not indicate the direction that the feature had on the likelihood of a purchase; ie, if the presence of the feature was more likely to lead to a purchase or less likely to lead to a purchase

Returned Values:

  • FEATURE_NAMES: name of feature
  • FEATURE_VALUES: relative importance of the feature
  • TIMESTAMP: unix timestamp of training run
  • MODEL_VERSION: version of platform that trained the model