Changelog

We’ve rolled out an upgrade to persona sets that makes them easier to interpret and more useful in practice.

Adding interpretation

Each persona now comes with:

• Summaries — each persona now includes a plain-English description in the Definition tab.
• Salient traits — the data points that most distinguish a persona from others in the set are featured in the Analysis tab and marked with a ⭐.
• Recommendations — practical engagement strategies now appear in a new Recommendations tab.

These additions appear in the Dashboard and API across new persona sets.

Learn more in the persona set documentation.

Efficacy reporting lets you measure how well your models are performing in the real world—comparing Faraday’s predictions to actual outcomes over time. Each report breaks down performance by decile, so you can quickly see how predictions line up with reality.

You can find efficacy reports directly in the Pipelines tab of the Dashboard for any pipeline that includes an eligible Outcome in its payload.

Developers can also access this data programmatically via the GET /scopes/:id/efficacy API endpoint.

View our documentation to learn more.

You can now receive email alerts when key Faraday resources finish building, refreshing, or encounter errors.

This feature introduces notification preferences under Settings → Notifications, where users can opt in to updates for specific resource events.

At launch, notifications will be turned off for most users and new users will also start with notifications turned off. You can opt in anytime from your settings, preferences are the same regardless of your plan.

See our notifications docs more information.

A regular vendor data refresh has been released. This update adds over 1 million new individuals to the Faraday Identity Graph and includes updated vendor-level datasets.

No issues have been observed so far, but please flag anything unexpected in your checks.

Trait appends can now fall back to broader levels of granularity when an individual match isn’t available in the Faraday Identity Graph (FIG).

Previously, traits were only appended at the individual level. Now, if we can’t identify a person directly, Faraday automatically resolves to the household or even regional level to ensure the record still receives relevant context.

For example, if a trait like age can’t be resolved for an individual, it will be approximated using the average age of the household associated with that address.

This update extends the same multi-resolution logic used in our predictive scoring to trait appends, improving overall coverage and consistency.

We’ve updated persona naming to remove automatically generated names and avatars.

Previously, persona sets included names like “Glamorous Grandma Gloria” to make results feel more human. But these labels could unintentionally introduce bias—implying that everyone in a persona shares the same gender, age, or lifestyle.

From now on, new persona sets will use neutral labels such as “Persona 1,” “Persona 2,” and so on. This provides a clean, unbiased foundation for your own interpretation and naming conventions.

Impact

• No API changes: names will simply appear as “Persona 1,” etc.
• Existing sets unaffected: older persona sets will retain their names until rebuilt.
• Optional customization: you can still generate descriptive names (e.g., using LLMs) based on the summary text if desired.

This change is live now and will roll out gradually to all accounts. Learn more in the persona set documentation.

In datasets console, you can now set an incremental column (e.g. updated_at, order_updated_at) on databases and data warehouses. This dramatically improves refresh efficiency by only processing rows that have changed since the last update.

When paired with a primary key, the refresh process works like this:

1. Select rows from the source where incremental_column is newer than the last run.
2. Merge (upsert) those rows into the dataset using primary_key.

This makes dataset refreshes faster and more efficient, especially for large tables.

We’ve expanded Credit Score Proxy, our predictive proxy for household financial reliability.

Now, in addition to the original 12-rank format (A1–E3), you can now use a simplified 5-category version for easier segmentation and reporting. Both formats are available in the Faraday Identity Graph and can be appended to your datasets like any other data point.

Learn more in our traits (or data points) docs.

Breaking change: We have dropped support for uploading datasets as multi-file ZIP or TAR archives due to unavoidable security issues. Raw CSV uploads remain supported, as well as single-file compression formats like .gz and .bz2 for batching and storage efficiency.

See our API docs for further information.

In outcome creation and editing, we moved the attrition cohort selection to the Advanced section. We determined that this field was being overutilized in unexpected ways, so this change should help create more accurate outcomes.

See https://faraday.ai/docs/abstractions/outcomes for more information.