Revamped product navigation
In the navigation menu, “Data” has been renamed to “Connections”.
As before, this is where you’ll go to add, configure and delete connections to external data sources.
In the navigation menu, “Models” has been split into “Blueprints” and “Models”.
Previously, blueprints were only accessible from the “Models” screen. This was where you would click to start creating a blueprint, and you had to click a particular model to view the blueprint it was trained on.
Now, blueprints and models are treated as two unique but linked assets, accessible from separate areas of the product.
Improved data-import experience
Available in Beta
Several updates have been made to the data-import experience based on user feedback. Here is an overview of the most notable ones:
The “Import settings” popup box has been replaced with a more spacious drawer, as seen below:
To provide more context and clarity around the data-import process, terminology has been updated and illustrations have been added.
You can now remove columns that were previously imported. (Previously, removing a column would have required you to import the entire table again.)
Now you can change the default way tables are imported before each prediction job is initiated.
To provide a clearer and more intuitive user experience, copy adjustments have been made across the board.
New Oracle connector
Available in Beta
Pecan can now import data from Oracle so it can be used to train models and provide the basis for predictions. In addition, you can export predictions from Pecan to your Oracle database.
To connect to Oracle, click Add connection under the “Connections” tab, select Oracle, and follow the on-screen instructions.
Export predictions to Firebase
Available in Beta
Pecan now supports sending predictions to Firebase, so teams can leverage predictions in the same platform they use to build and ship their applications.
To connect to your Firebase account, click Add connection under the “Connections” tab, select Firebase, and follow the on-screen instructions.
Optimize LTV models by aggregation level
Available in Beta
For Lifetime Value models, you can now optimize your model according to a defined aggregation level. Predictions will still be generated on the customer level, but your model will optimize itself based on the chosen metric (e.g. Weighted MAPE) for the defined aggregation level (e.g. “Campaign”, “Platform”).
This better supports marketing optimization and campaign management efforts. To choose an aggregation level:
Make sure the relevant column (or columns) is in your Entity table.
Open the relevant blueprint, and click Train model to open the “Model Settings” drawer.
Under “Select optimization metric”:
Indicate your model type (if it’s not already selected).
Select the performance metric by which you want to optimize the model (e.g. WMAPE).
Select the column (or columns) that defines the aggregation you want to optimize the model for.
If you have questions about this feature or how to use it, feel free to reach out to your customer support manager.