Data Retention

Nova keeps the data needed to score an application for a limited period, then removes application input data automatically. You can also request deletion for one application or one tenant through the API.

​

Automatic cleanup

​

Stored results

​

Background processing

Nova uses background workers to process scoring, deletion, scheduling, and webhook delivery. Worker records keep operational metadata such as IDs, retry state, status, and timing. They don’t include resumes, application answers, scores, assessments, webhook bodies, partner response bodies, or presigned resume URLs.

​

Deletion requests

Use programmatic deletion when your customer leaves, a tenant asks you to delete its data, or an application must be removed before automatic cleanup runs. Deletion requests are asynchronous. The API returns 202 Accepted with a deletion request ID, then Nova processes the deletion in the background. Poll the deletion request endpoint for completed or failed. Nova uses your API key to scope every deletion to the authenticated partner and environment. Application deletion removes active data linked to the specified job and application. Tenant deletion removes active tenant-scoped Embed data. Unknown tenant or application IDs are accepted and processed without revealing whether matching data existed. Deletion request responses don’t expose internal cleanup counts, storage details, raw tenant IDs, job IDs, or application IDs. Start with Deletion Requests, or open the endpoint docs for application deletion, tenant deletion, and status polling.

​

Cleanup schedule

Cleanup runs daily. It removes expired scoring inputs, stale resume references on expired scoring jobs, expired question sets, and eligible webhook delivery records.

​

Retained copies

Backups and retained object copies created under the current retention policy are configured to age out within 30 days. Copies created before the current policy may age out under their original schedule. Operational and security logs may be kept separately for service reliability and security evidence. They are designed not to contain resumes, application answers, scores, assessment text, webhook payload bodies, partner response bodies, or AI prompts. Logs and telemetry may keep routing metadata, status codes, timing, model names, token counts, and safe error categories. Current Embed sub-processors are listed on our Embed API Sub-processors page.