Book a demo

DiscoverThe data platform

One version of every number

Connect the databases you already run, agree what each number means once, and decide who sees which rows. Every dashboard, chat answer and workflow then works from the same definitions, so meetings stop arguing about whose figure is right.

Certified metrics

Agree what a number means, once

A metric gives a measure a name, a meaning and an owner, so everyone means the same thing by "net revenue". Certified metrics are the ones the assistant reaches for first, and a new measure that duplicates one is offered the existing one instead.

  • Each metric has an owner, a description for people and for the assistant, and its own filter
  • Rename it and nothing breaks: dashboards point at the metric, not its name
  • A model is certified as the one to build on, and flagged when it changes

The assistant beside the editor

Model your data by describing it

Say what you need, "a sales model from orders and lines, revenue net of discount, each rep sees their own region", and the assistant drafts it into the editor. Every change is ringed and waits for you: nothing is saved until you save it.

  • A modelling expert reviews your models and drafts the fixes
  • Help moving your Power BI measures across
  • Drafts are visible only to you until saved

Time intelligence

Year to date and last year, without writing SQL

Year to date, growth on last year, rolling twelve months, closing balances and share of total are one short formula each, and they work in any chart, table or KPI. Your fiscal year is set once on the model.

  • Recipes for the measures finance asks for most
  • Formulas are checked as you type, and can never reach rows you may not see
  • Ranks, top N and percent of parent for the rest

Row access

Everyone sees their rows, and you can prove it

Rules by role and by people's attributes, such as their region or company, decide which rows each person sees, for dashboards, chat, workflows and embeds alike. Look at the model as any person before you save, and someone with no rule sees nothing.

  • "View as them", with live row counts and a sample of their rows
  • More roles never mean more rows: forgetting one hides data
  • Sensitive models log every read: who, when and with which filters

Explores and dashboards

Dashboards people trust, built in a few steps

Drag certified measures and groups onto a chart, pick from 26 chart types, and lay them out on a dashboard with filters that reach every tile. Or ask the assistant to draft the dashboard, and check each tile before you save it.

  • Certified fields marked with a tick, so builders reach for the right ones
  • Clicking a mark filters the whole dashboard
  • Publish to View, email on a schedule or share a link when it is ready

Lineage and impact

Change things without breaking them

Before you rename, rework or delete anything, see exactly what would break, what would show different values and what is untouched, across every model, explore and dashboard. Lineage shows where any figure comes from and what uses it.

  • "What saving changes", shown before every save
  • From connection to table, model, explore and dashboard
  • What you cannot open is counted, never named

Warehouse and synthesis

A home for the data that has nowhere to go

Each workspace has its own fast warehouse. Drop in a spreadsheet or a file and model it like any database, or keep a copy of a slow system so reports never load it. Need test data? Describe it, and a generator makes it in a sandbox, marked synthetic everywhere.

  • CSV, Excel, JSON or Parquet, as a new table or added to one
  • Synthetic data is drafted as a plan and a script you can read first
  • The model never sees a real row of the table it extends

Connections

Use the databases you already have

Connect to 36 databases, warehouses and hosted services where they are, without moving your data. Reach private networks through SSH, sign in with Microsoft Entra, and keep credentials sealed: they are never shown again once saved.

  • 17 database engines, from SQL Server and Oracle to Snowflake and BigQuery
  • Azure Log Analytics queried directly, in its own language
  • A read-only database user is recommended, and the form says so

Also in Discover

What keeps the data layer healthy as it grows.

  • Model reviewA score for how reusable your models are, with a Fix button that drafts each fix.
  • Relationships you can seeFacts in the middle, lookups around them; drag a field to propose a link, and it is checked.
  • Profile and cacheEach column's blanks and spread, and a copy of slow models' rows kept on a schedule.
  • Devices and sensorsReadings from devices, shaped into tables in your database, with alerts that watch them.
  • Hidden and masked fieldsMark a field personal or secret, and decide which roles may use it.
  • Write back from an exploreLet chosen roles correct a row in the database, from the table they are reading.
"Give our customers the opportunity to play with their own data, without breaking anything."
Kyle Pillay, SOC Manager, DatacentrixCustomer stories →
Next · Orchestrate
Processes that run themselves→ View
Where everyone reads what you build→