The attributes endpoint provides hundreds of attributes that describe a user’s cash flow and financial health, including their past and expected income and bills, loan repayment history, balances, etc.
These attributes are based on models we’re developing to predict outcomes including a user’s likelihood to pay back a cash advance or loan. The format of this endpoint is easy to consume and allows data scientists and developers to join with existing (non-Pave) attribute sets.
You can read more about the endpoint in our documentation.
Using the GET attributes endpoint, you’re able to retrieve ready-made features to:
- Train or backtest your models, including credit risk models
- Use in production models for real-time predictions
- Build customer segmentation by comparing the historical and expected future behavior of each end-user
Here are some FAQs to get you started!
Can I generate attributes for a specific date in the past?
Yes. You can select a specific date from which the features are built by passing in the date parameter when calling the attributes endpoint. For example, “income_past_7d” will be the amount of income for the 7 days period ending on but excluding that date. The transactions that occurred on that date are excluded from the computation of attributes. This allows you to avoid issues of data leakage.
For example, if you would like to train a credit scoring model, you want to use features that precede the date of the loan, since this reflects your knowledge of the user’s behavior prior to extending them a loan. Therefore, the attributes endpoint excludes transactions that happened on the date to allow you to pass in the date of the outcome you would like to predict.
Example API call –
How often are the attributes refreshed/updated?
The attributes returned by the endpoint are computed when you call the endpoint. Therefore, they represent our latest capability in cleaning and tagging transactions. This also means that the values returned may change between API calls as we improve our tagging of transactions.
For example, the value of “income_past_7d” using the same date parameter may change between different API calls if we can detect more income for the user. The value may also change if there are changes in the user’s transactions (e.g., some transactions being deleted or uploaded).
How many attributes are returned by this endpoint?
We currently have 124 attributes and are continuously adding new ones!
I would like to access the attributes for a large number of my users without calling the endpoint multiple times. How do I do that?
We are building infrastructure to share the attributes through Snowflake secure data sharing. This means you would be able to retrieve the attributes for all your users in a table format. If interested, please reach out to us for more details.
You can find the documentation here – https://docs.pave.dev/attributes