Answers that show their working
Ask which brand lost the most organic traffic last month. Get the answer, the chart, and the exact query that produced both.
The problem
An analytics chat that hands you a confident paragraph and no way to check it is worse than no chat at all. You cannot ship a number you cannot reproduce, so you end up rebuilding the query by hand anyway.
What we mean
The model picks the question; the semantic layer answers it. Every response carries the query it ran: measures, dimensions, filters, window. You can read it, change it, and save it as a chart.
0
answers without the query that produced them
How it works
Start to answer
- 01
Ask
Plain English, with the brand and the period however you would say them out loud.
- 02
It picks a query
From the schema only. It cannot invent a metric that does not exist.
- 03
You see both
The answer, the chart, and the query underneath it.
- 04
Keep it
Pin it to a dashboard or export the whole thread as a report.
Before and after
What actually changes
Before
Answering 'and what about mobile?' means rebuilding the report with one more filter.
After
You ask it. The chart updates and the query shows the filter.
Before
A screenshot in Slack, and a week later nobody can say what date range it covered.
After
An exported report with the window and the query stated on the page.
Auditable by default
The query is not hidden behind a toggle.
- Measures, dimensions, filters and window, shown
- Editable in place
- Saveable as a chart you own
Grounded in your data
It answers from the warehouse, never from memory.
- No answer without a query behind it
- Says when data is missing rather than smoothing over it
- Scoped to the brands in your workspace
Made to be kept
A thread is a document, not a scrollback.
- Every chat is saved and searchable
- Charts pin straight to dashboards
- Word and PDF export
Where the model is not allowed to go
It does not write SQL against the warehouse directly, and it does not decide what a metric means. It selects from declared measures and dimensions. That constraint is the reason the answers can be trusted, and the reason a wrong answer is a wrong question rather than a wrong number.
In short
The question you would have spent an afternoon on becomes a sentence, and the answer arrives with its receipts attached.
More features
One metric schema
Every source normalises into one schema on the way in, so comparing two brands is valid by construction rather than by whoever built the spreadsheet remembering to make it so.
Portfolio view
The brand index does not average your portfolio into a single score nobody can act on. It sorts by movement and flags what needs you.
Charts & dashboards
Every chart is a saved query rather than a picture, so a dashboard is a set of live questions instead of a set of stale images.
See the whole brand
Search, analytics, ads, revenue, CRM and support, pulled into one view per brand and kept current. Connect the first source in a few minutes.