Data profiling
Know your data, and when it changes.
Column-level statistics on every run, trends over time, and machine-learning thresholds that flag anomalies without hand-written rules.
Deep insights into dataNull counts, uniques, min/max, averages and distributions per column, plus custom rules.
Track data changesTrend graphs across scheduled profiling runs, with granular history management.
Anomaly detectionForecast-based thresholds, new-value detection and alerts when a column deviates.
Quality score, freshness and row countObject-level health with alerts for the responsible users, and a history for every rule.

Deep insights into data
Every profiled column gets the standard statistics out of the box, and your own rules on top.
- Column-level statistics out of the box: null counts, unique values, min/max, averages, sum, string lengths and value distributions by value, range or prefix
- Custom profiling rules per column for tailored analysis beyond the defaults
- Profiling results directly in the catalog and used to generate data validation suggestions

Track data changes
Profiling on a schedule turns a snapshot into a history, so a shift in a column is visible the day it happens.
- Historical trend graphs showing how column-level metrics evolve across every profiling run
- Schedule profiling tasks by interval or CRON expression to keep metrics continuously fresh
- Backfill past days from a timestamp column, so a newly profiled table starts with a history instead of a single point
- Granular history management: exclude individual executions or values from analysis, or mark a failed run as expected, without losing the underlying data

Anomaly detection
Instead of fixed limits, thresholds adapt to how each column actually behaves, and alert when it stops.
- Forecast-based thresholds that follow each column’s own behaviour, including seasonal patterns, so a recurring weekly peak is not an anomaly
- New-value detection: a value never seen in a column before is flagged the first time it appears
- Automatic or manual limits: let the model set the band, add a minimum, or define a fixed percentage band yourself
- Configurable alerts per profiling object, delivered by email or webhook

Quality score, freshness and row count
Each profiled object carries its own health summary, so a table that goes stale, shrinks or drops in quality is noticed by the people who own it.
- Overall quality score per object, tracked over time, with alerts to the responsible users when it changes
- Freshness and row count checked on every run, with deviations and anomalies flagged against the object’s own history
- Rule-level overview and history: see every profiling rule on the object, its current value and how it has developed run by run

See it on your own data.
SelectZero runs inside your own environment, with your access rules in place and no data leaving it. Try the sandbox, or book a demo.
