Data preparation with data science in mind. Narrator makes creating tables with the right features super easy, so you can spend all your time on analysis and writing algorithms.
Show MeDataset allows you to create any table you need instantly so you can prepare the data without having to join a bunch of tables in memory in your notebook.
Once you create your dataset, materialize your feature aggregation and drag it into your notebook to run your algorithms.
Add any dimension, computation, or aggregation to your dataset in minutes and it’ll show up automatically in your notebook.
Get time-based features like “the last call before they signed up” easily because everything in Narrator is based around time and occurrence.
Deal with billions of rows easily because Narrator’s processing happens on your data warehouse. Materialize only aggregated feature counts so there’s no need for Spark.
Once you get an algorithm working, instead of the data engineering team having to go through your notebook to try and recreate your work, you can now give them the SQL query generated by Narrator and your algorithm and get it in production in minutes.
Use Narrator’s Analyze Button to test if something is impactful via Naive Bayes to see if there’s any signal before spending your time diving deeper in a notebook. With one click any feature can be evaluated for you to see if it is predictive.
Narrator intelligently segments your data based on trends, applies intelligent dimensional reduction, checks for consistency to ensure time variance is not an issue, checks for statistical significance, and even simulates the impact of changing the feature.
Narrator is the perfect tool for data modeling at scale. Using the activity schema allows speed, simplicity, and control.
Narrator’s activity schema based approach makes data lineage transparent and easy to follow. All tables, metrics, and plots are built on top of activities.
Narrator’s activity schema approach removes the need for a traditional data dictionary.