01
Native auditing per source
Each database is checked for what tends to go wrong in it, not against one generic list applied to all.
- Connectors for eight kinds of source
- Read-only queries only
- Scheduled audits per connection
Auditing · Governance · Privacy (LGPD)
Auditing and governance for your entire data ecosystem
Diagnostics for relational databases, warehouses, CDPs, caches, graphs and vector stores, deliberated by a council of specialists, with catalog, policies and data quality in one place.
8
kinds of source
9
council specialists
P0–P3
severity per finding
2FA
required for everyone
Platform
Six capabilities, one workflow
From the first check to the remediation report, with data governance alongside the health of the databases that hold the data.
02
Findings that prioritise themselves
Each finding carries severity, magnitude, blast radius and trend, so you can decide what to fix first.
- Severity P0 to P3
- Evidence and remediation SQL
- Rollback plan per finding
03
Deliberative council
Nine specialists review each finding through their own lens, from performance to privacy and sector regulation.
- Deterministic deliberation, no LLM
- Contested findings go back to the council chair
- The same finding always gets the same verdict
04
Catalog and privacy
The inventory of tables and columns is harvested from the source itself, with suggested personal-data classification.
- Personal and sensitive columns flagged
- Domains with an owner and a steward
- Asset certification
05
Data quality
Declarative rules written by the data steward and run by the same engine as the audit.
- Completeness, uniqueness and freshness
- Reconciliation between two sources
- A critical failure downgrades certification
06
Policy as code
Naming, classification, masking, ownership and retention rules evaluated on every harvest.
- Justified, time-boxed exceptions
- Approval recorded under the approver’s name
- Compliance rate per asset
How it works
Five steps from connection to fix
None of them changes the audited source: Optimize reads, deliberates and recommends; your team applies.
- 01
Connect
Register the source with a read-only user. The credential stays in Optimize and is never shown again.
- 02
Diagnose
The source’s checks run on demand or on a schedule and become findings with evidence.
- 03
Deliberate
The council confirms, contests or dismisses each finding, and the synthesis explains why.
- 04
Govern
Catalog, policies and quality rules show who is accountable for each piece of data.
- 05
Remediate
The executive and technical report brings the remediation SQL and the rollback plan.
Supported sources
What is checked in each one
The checks follow the way each technology fails. Catalog and quality rules are available on the SQL sources with metadata harvesting.
| Source | What Optimize checks | Catalog and quality |
|---|---|---|
| PostgreSQL | Vacuum and dead tuples, indexes, locks, connections, SSL and roles, WAL and backup, version | Yes |
| Snowflake | Auto-suspend and credit usage, failed queries, perimeter segregation, masking, privileges | Yes |
| BigQuery | Large unpartitioned tables, views with SELECT *, high-cost queries | — |
| Salesforce Data Cloud | Data stream health and freshness, personal data modelled in DMOs | Yes |
| ClickHouse | Parts per partition, pending mutations, replication queue | — |
| Redis | Memory fragmentation, maxmemory and eviction policy, keys without TTL, slowlog | — |
| Neo4j | Integrity constraints, indexes, supernodes | — |
| Qdrant | Collection status, payload index, very large collections | — |
FAQ
What people ask before connecting
Does Optimize change anything in the audited databases?
No. The checks only run read queries against the source’s catalogs, statistics and settings. Fixes appear as recommended SQL in the report, for your team to review and apply.
Does the council use generative AI?
No. Deliberation is deterministic: each specialist applies its own rules to the finding, and the same finding always gets the same verdict. None of your data is sent to a language model.
Where are the source credentials kept?
In Optimize’s metadata database. Passwords, tokens and keys are never returned by the API or shown in the interface; when a saved connection is tested, the server itself uses the stored credential.
How does access to Optimize work?
Each person has an account created by an administrator, with mandatory two-factor authentication. The ADMIN, DBA, ENGINEER and VIEWER roles define what each person can see and do, and sign-ins and account changes are logged.
Which sources have catalog and quality rules?
PostgreSQL and Snowflake. On both, Optimize harvests tables and columns, suggests the personal-data classification and runs quality rules. The other sources get the audit checks.
Ready to audit your sources?
Sign in with your account and the code from your authenticator app.
Go to the platform