Accepted data
Accepted usage is the uncompressed source size of records that pass validation and deduplication. Flarehawk records usage only after the required storage writes succeed. The following contribute zero billable usage:- Rejected records
- Duplicate batch deliveries
- Retry attempts for an already accepted batch
- Batches that fail ClickHouse persistence
- Investigate or Enterprise batches that fail required R2 archive durability
AI investigation cases
An AI case is counted once when an incident-backed or manually initiated investigation first produces a substantive successful result. A failed attempt does not count. Retrying that attempt, reanalyzing the same incident, asking a follow-up, or making additional internal model calls does not create another billable case.Viewing usage
Open your organization and select Billing. The billing view shows:- Accepted GB and AI cases in the current subscription period
- Data and AI allowance percentages
- Forecast end-of-period usage
- Estimated current-period invoice
- Retention and independently configured caps
- Pending plan changes and renewal date
Usage notifications
Flarehawk sends period-scoped notifications when either meter crosses 80%, 90%, and 100% of its allowance or configured cap. Each threshold is sent once per cap revision and resets with the next subscription period.Data overage
Paid plans continue ingesting by default. Overage uses accepted data above the included allowance:
Stripe receives cumulative decimal-megabyte totals through an idempotent meter. Dashboard estimates and invoices use the same versioned catalog.
AI overage
Free supports no data or AI overage.
Independent hard caps
Paid plans default to continuous ingestion and investigation with overage billing. An organization admin or owner can independently enable:- Data cap — rejects new batches with a retryable plan-limit response. Existing data remains available.
- AI cap — stops new billable investigations. Ingestion and non-AI detections continue.
Plan changes
- Upgrades from Detect to Investigate apply immediately after a server-side Stripe preview and are prorated.
- Downgrades, cancellation, and transitions to Free apply at renewal through a subscription schedule.
- Scheduled changes can be canceled before their effective date.
- Usage already accepted in the period remains part of the authoritative period total.
Retained data
Investigate retains canonical searchable logs for one year. Enterprise contracts retain canonical silver Parquet logs for one to five years, with the first 30 days in Standard object storage and later data eligible for Infrequent Access. Normal bounded cold search is included; bulk restore, export, backfill, and extended provider-original raw retention are quoted separately.Frequently asked questions
When does usage reset?
When does usage reset?
Usage resets at the start of the next Stripe subscription period. It does not reset on the first day of the calendar month unless that is your subscription boundary.
Can a retry be billed twice?
Can a retry be billed twice?
No. Accepted data has one logical identity per organization, batch, dataset, and revision; its billing period is immutable payload on that identity. A retry that crosses a subscription boundary therefore cannot become a second accepted-usage row. Stripe meter events also have deterministic identifiers, so delivery and webhook retries remain idempotent.
Does a follow-up question consume another AI case?
Does a follow-up question consume another AI case?
No. Follow-ups and internal model calls remain part of the original investigation case.
What happens when only the AI cap is reached?
What happens when only the AI cap is reached?
New AI investigations pause, but telemetry ingestion and detections continue. The data cap is independent.