Data Profiling

Onboarded data sources with KPIs, dimensions, and domain knowledge that power every downstream feature.

What Datasets is

Datasets are onboarded data sources with KPIs, dimensions, and domain knowledge that power every downstream feature. A Dataset is the operational definition of what DataGenie monitors autonomously — its KPIs, dimensions, time anchor, granularity, and anomaly detection configuration. Everything else in the product (Top Stories, Wisdom, Explorer, Dashboards) reads from this definition.

The fastest way to create a Dataset is through GO — describe your business problem in natural language and GO generates the Blueprint for you.

How DataGenie profiles your data

When you point DataGenie at a data source, its Query Engine automatically scans, describes, and maps your data - schemas, tables, columns, and the relationships between them - in plain language. You get instant clarity on even a complex dataset, and you stay in control: edit or add descriptions, relationships, and tribal knowledge any time.

Connect your database

Point DataGenie at your source. It immediately discovers the structure - for an e-commerce dataset, that might be a schema, its tables, dozens of columns, and the relationships between them.

See your data described automatically

DataGenie detects and describes every element in plain words - orders, sessions, marketing spend, daily stock - and shows how connected each table is, so you know which sources stand alone and which link together.

Explore any table

Open a table for instant context: what one row means, the insights that matter (for example, each session produces at most one order), and the gotchas an analyst needs - like marketplace vs retail orders being read differently, so nothing gets double-counted. Each column carries a description and sample values.

Visualize and refine relationships

DataGenie maps how tables connect, with a connection confidence and plain-language guidance (for example, "start from Sessions when measuring conversion - sessions without orders still count"). Edit, remove, or add relationships right from this view.

Add your own knowledge

Add or update descriptions, relationships, and tribal knowledge directly. DataGenie profiles your data deeply and builds context automatically - you fine-tune it as you go.

When to use Datasets

You need to fine-tune KPIs, dimensions, or SQL transforms after GO onboarding
You want to configure or tune anomaly detection for specific KPIs (see Multi Yhat)
You want to set up multi-source monitoring with Nirvana / Hyper Connected
You want to change processing cadence, impact score weights, or third-party alert channels

What you get

Essential details

Dataset name, connection parameters, time anchor, SQL transformation, and tenant association.

Regular processing

Processing range, schedule (Daily / Weekly / Monthly), and thresholds.

Build your insights

KPIs, dimensions, computed attributes, and anomaly detection configuration — the autonomous monitoring surface.

Customize insights

Impact score weights and Filter Presets for Top Stories.

Third-party alerts

Slack, Teams, Email, and JIRA channels for alert delivery.

Nirvana datasets

Combine independently-onboarded sources into a virtual multi-source dataset — no ETL.

How it works

Set the Time Anchor

Define which column represents the date/time dimension. This is the reference for all period comparisons and anomaly detection.

Define Business KPIs

Add the metrics DataGenie should monitor. Specify aggregation (Sum, Average, Count) and any SQL transformations.

Configure Columns & Dimensions

Mark which columns are dimensions and set their cardinality.

Set the Processing Schedule

Choose how often DataGenie re-processes the dataset — Daily, Weekly, or Monthly.

Configure Anomaly Detection

Assign an AD Group to your KPIs. See the Anomaly Detection guide for Multi Yhat, seasonality handling, and Nirvana inheritance.

(Optional) Compose a Nirvana dataset

Combine multiple independently-onboarded sources into a virtual multi-source dataset. Nirvana datasets inherit default AD Groups from their components.