The analyst that shows its working.

Imago connects to the databases your team already runs and answers questions in plain words. Every answer comes with the chart, the table and the query behind it, so you can check a number before you repeat it.

Get started Book a demo

Question: Why are returns up in Central Europe?

  1. Planned 2 queries
  2. Read shop-prod, read-only
  3. Checked 4 figures against the results

Imago’s answer: Returns in Central Europe jumped from of orders in week 35 to in week 36, and have stayed above 12% for . Footwear is of the returns since then.

Written by Imago from 2 queries on shop-prod (PostgreSQL)

Show the 2 queries

1. The weekly return rate

SELECT date_part('week', o.ordered_at) AS week,
       COUNT(DISTINCT r.order_id) * 1.0
         / COUNT(DISTINCT o.id) AS return_rate
FROM orders o
LEFT JOIN returns r ON r.order_id = o.id
WHERE o.region = 'Central Europe'
  AND o.ordered_at >= DATE '2026-07-27'
GROUP BY 1
ORDER BY 1;

2. Returns by category since week 36

SELECT p.category,
       COUNT(*) AS returns,
       COUNT(*) * 1.0
         / SUM(COUNT(*)) OVER () AS share
FROM returns r
JOIN order_lines l ON l.id = r.order_line_id
JOIN products p ON p.id = l.product_id
WHERE r.region = 'Central Europe'
  AND r.returned_at >= DATE '2026-08-31'
GROUP BY p.category
ORDER BY returns DESC;

Return rate, Central Europe

Share of orders returned, by week

Returns since week 36, by category

CategoryReturnsShare
Footwear3,90464%
Outerwear1,15919%
Knitwear67111%
Accessories3666%

Square corners mean your database reported it: charts, tables and every value in them.

Petal corners mean Imago wrote it: answers, drafted descriptions and proposed actions.

Point at or tap any figure in the answer to see where it came from.

From a database login to the board your team opens each morning

Teams take Imago up in this order. Each step works on its own, and each one gives the next more to go on.

  1. Connect

    Connect the databases you already run.

    A three-step wizard tests the login before anything is saved, lists the schemas it can read and tells you whether that login could write, so you can give Imago a safer one. Every query then runs read-only, with a time limit.

    Works with PostgreSQL, MySQL, Snowflake, BigQuery, Amazon Redshift, Databricks, Teradata, Apache Spark, SQLite and DuckDB.

    Test the connection

    • Connected to shop-prod as analyst
    • Found 3 schemas and 48 tables
    • This login can write to 6 tables. Imago only reads, but a read-only login is safer.

    Nothing is saved until the test passes.

  2. Model

    Imago learns your schema. You approve what it learns.

    It samples each table and drafts a description of every column. Your team accepts, edits or rejects each one, and every answer reads from what you approved. It also finds joins your schema never declared, and measures them before it asks.

    returns.reason Draft

    Why the customer sent the item back, chosen from a fixed list at the returns desk. Empty in 2% of sampled rows.

    Drafted by Imago from 1,000 sampled rows

  3. Ask

    Ask, then keep asking.

    Follow-ups carry the context. “And the year before?” becomes a complete question before anything runs, and “show it as bars” redraws the last answer without running the query again. When two sources hold a table of the same name, Imago asks which one you mean.

    And the year before?

    Read as

    Weekly return rate in Central Europe, weeks 31 to 46 of 2025

  4. Act

    Imago flags what changed before anyone asks.

    Each morning it reruns the queries behind your shared dashboards and writes a short briefing. The words are Imago’s; every number in them is computed from the results. A finding becomes an action your team investigates or dismisses, and every decision is kept.

    12.9%

    Return rate, Central Europe, week 46

    ▲ Up 6.8 points on week 35

    Returns in Central Europe have stayed above 12% for eleven weeks, about twice the rate before week 36.

    From this morning’s briefing

  5. Build

    Describe a board in a sentence.

    Imago’s agents plan the KPIs, write and run each query, lay out the sections and check the result before you see it. Refine it in chat or drag tiles where you want them. Its filters re-run every query, so a filtered number is a real one.

    A returns board for Central Europe, by category and week

    6,100

    Returns since week 36

    12.9%

    Return rate, week 46

    Returns since week 36, by category

    • Footwear3,904
    • Outerwear1,159
    • Knitwear671
    • Accessories366

    Footwear is 64% of the returns since week 36.

Your data stays where it is.

Imago reads your databases where they live and treats every account as if it were the only one.

Read-only, whatever the login
Every query runs on its own, in a read-only transaction, with a time limit. Statements that could write, or call into the server, are refused before they run.
Accounts walled off
One account’s sources, dashboards and conversations can’t be reached from another, and tests check it on every change.
Sign in the way you already do
Single sign-on with Okta, Entra ID or Google, people provisioned with SCIM, and two-step sign-in when you require it.
Permissions by capability
Decide who can view, ask, build dashboards, approve models or manage connections, not only who is an administrator.
Credentials sealed
Database passwords and keys are encrypted at rest and never sent back to the browser.
On your own servers
Run Imago inside your network with open-weight models, so no question and no row leaves it.

Bring the question your team keeps asking.

Sign in to your workspace, connect a database and ask it something. The answer comes with its working.

Get started No workspace yet? Book a demo