ML Models Overview

ML Models are MorphL's core product offering. Using Machine Learning, you can predict users' behaviour in websites or applications and personalize their experience based on their needs.

MorphL allows you to connect your existing analytics tools (called "Data Sources") and feed data into ML models (ex. Shopping Stage Completeness, User Search Intent, Product Recommendations, etc.). The continuous stream of predictions generated by each model is then "translated" into different formats for use by other tools (called "Destinations") and pushed into those tools.

What's a ML Model?

Machine Learning is a subset of artificial intelligence, a method of training algorithms such that they can learn how to make decisions. Machine learning relies on working with small to large data-sets by examining and comparing the data to find common patterns and explore nuances. Compared to rule-based systems which rely on encoding knowledge as fixed rules, a ML model can adapt to new situations by creating new rules.

The majority of MorphL ML models use supervised learning. A notable exception is our Customers segmentation model which generates clusters of users based on their activity / actions (unsupervised learning).

How it works

MorphL is a single-stop shop for ML models tailored for eCommerce businesses. In contrast to generic analytics platforms (ex. Mixpanel, Kissmetrics) and generic ML platforms (ex. Amazon Web Services, Google Cloud or Azure), MorphL uses specific eCommerce data and provides specific eCommerce predictions.

The eCommerce data used by the MorphL models includes (and it’s not limited to) the following:

  • User & browser characteristics (ex. device, city / country, source / medium, interests).
  • Browsed products and products lists, including products characteristics
  • Search activity
  • Shopping cart and checkout activity
  • Wishlist / favorite products activity
  • Orders history
  • Email activity
  • Push notifications activity
  • Other features derived from above primary features (such as orders seasonality, standard deviation of purchases dates, average order value, etc.), which are generated inside the MorphL platform.

From the MorphL Dashboard, you can connect a data source (or more than one!). Once a source has been connected, the ingested data can be used to train multiple models.

A model can be attached to a single data source at a time, but our models are compatible with different source types. For example, our Shopping Stage Completeness model can be trained on Google Analytics, Google Analytics 360 or BigQuery data.

ML Models catalog

ML Models labels

Each ML model has a unique label that is used when calling the MorphL API to retrieve predictions or by destinations for saving predictions.

ML Model Label
Shopping Stage Completeness shopping_stage
Cart Abandonment cart_abandonment
Customers LTV customers_ltv
User Search Intent user_search_intent
Customers Segmentation customers_segmentation
Personalized Recommendations personalized_recommendations
Related Products related_products
Frequently Bought Together bought_together
Churning Users churning_users
Forecasting forecasting
Next Order Value next_order_value