Data quality monitoring FAQ

Straight answers to the questions teams ask most when evaluating data quality monitoring. For background, start with What is data quality monitoring? or browse the glossary.

How do I monitor data quality without exposing raw data to a vendor?

Run a monitoring agent inside your own environment instead of shipping rows to an outside service. AIMO's agent runs as an open Docker image in your network, connects to your database locally, and sends out only aggregates and metadata: counts, monitor values per time block, and analysis payloads. Bulk raw rows never leave your environment, so there is no firewall opening and no third-party access to your database.

Can AIMO run entirely on-premises?

Yes. The standard setup keeps the agent in your network and the coordinating service in AIMO's cloud, so only aggregates and metadata cross your perimeter. Larger organizations whose policy allows nothing outbound can deploy the whole platform on-prem, the cloud side included, and point it at an LLM they supply, so nothing leaves at all. Terms are shaped to the deployment and volume rather than the per-table rate; talk to us.

What is the difference between data quality and data observability?

Data quality is about whether the content of your tables is correct: completeness, freshness, valid ranges, consistent relationships. Data observability is the broader discipline of understanding the health of your whole data system, often including pipeline lineage and infrastructure. AIMO focuses on table-level data quality, with outlier detection learned from each monitor's own history.

Can AI generate data quality checks automatically?

Yes. AIMO analyses your schema and a statistical profile of your data, then uses an LLM to write the monitors: which checks, on which columns, with which valid ranges and value sets. Every generated monitor starts active, so you get broad coverage in minutes and only decide what to switch off. Alert thresholds are never hand-set; the detector owns them. See monitors.

How does AIMO detect outliers and anomalies?

For every monitor and time block AIMO rolls an expected value forward from the monitor's own history, including its daily and weekly rhythm and the momentum of the last few blocks. Readings that leave the band of ordinary variation build suspicion block by block, and readings that return drain it. When suspicion crosses the alert level, set by your account's alert budget, the blocks become one episode and one alert. Nothing is trained, and there are no per-monitor thresholds to tune. See outlier detection.

Which databases can I monitor?

The agent connects through SQLAlchemy and ships first-class support for PostgreSQL 10 or later (including Databricks Lakebase), MySQL, Microsoft SQL Server, Snowflake, BigQuery, Databricks, Amazon Redshift, DuckDB, and SQLite. Each connection accepts either individual fields (host, user, …) or a complete SQLAlchemy URL. See getting started.

Is AI-based data monitoring secure and GDPR-compliant?

Security is built in. There is no path from the AIMO cloud into your database: the agent connects outbound only, runs every query beside your data, and sends out aggregates and metadata rather than raw rows. Database credentials are encrypted on the agent with a passphrase that never leaves your environment; only the ciphertext is stored, so AIMO can never read or use them directly, and you should give the agent a read-only database role. Each agent authenticates with an asymmetric keypair (Ed25519 or RSA) and short-lived JWTs; human sign-in uses passkeys (WebAuthn), with no passwords. AIMO is EU-based and GDPR-ready, operated from Finland by Aimodata Oy. See security and the legal documents.

How do I get alerted when something breaks?

Alerts are delivered to email and Slack, one per episode rather than one per bad reading. You configure destinations and routing yourself in the UI, with severity (info, warning, error, critical) driving where each alert lands. How often alerts may come at all is one account-wide setting, the alert budget, and the monitors that built suspicion without reaching it are listed as held silent so nothing is lost. Microsoft Teams, Linear and generic webhook destinations can be set up and test-pinged from the same screen.

How long does setup take?

Minutes, not months. You run the Docker image, register in the UI, connect a database, analyse it and switch on the tables to monitor. AIMO generates the monitors, prices each table's history and begins monitoring; the backfill time scales with your data size. See getting started.

How much does data quality monitoring with AIMO cost?

Your first month includes up to three monitored tables at no charge. After that it is €10 per monitored table per month (excl. tax), with no per-seat fees and no platform fee. Add or remove tables in the UI; removing a table stops its billing at the next cycle. See pricing.