Platform · How it works
Illustrative traceWhat happens between a question and an answer.
One question, traced through QuerySafe Intelligence: “Which customers have reduced their purchasing significantly over the last 90 days?”
01 · Before any question
Connect
An administrator connects QuerySafe to the BigQuery datasets the organization wants it to use.
- › project · demo-analytics
- › datasets · sales, crm, catalog
- › access · read
02 · 0.3 s
Understand
The question is interpreted: what is being measured, over what period, compared with what, for whom.
- › intent · purchasing decline
- › entity · customer
- › window · 90d vs prior 90d
- › threshold · significant = ≥ 40%
03 · 0.6 s
Retrieve
QuerySafe identifies which tables and fields hold the answer, using what it has learned about your data.
- › sales.orders
- › sales.order_items
- › crm.customers
04 · 1.4 s
Reason
The analysis runs against the relevant data. Results are compared, grouped and checked for what stands out.
- › 1,284,902 rows scanned
- › 4,812 customers compared
- › 327 matched
05 · 0.5 s
Explain
Numbers become an answer: a plain-language summary, the key figures, a chart and the sources used.
- › summary · 2 sentences
- › metrics · 4
- › chart · weekly orders
- › sources · 3 tables
06 · Next
Explore
The conversation continues. Follow-up questions reuse the context of the previous answer.
- › “Which of these were top-20 accounts last year?”
Throughout
Boundaries apply at every step.
What a person can ask about depends on what they are allowed to see. Where Agent capabilities are enabled, actions pass through permission checks and, where appropriate, human approval before anything is executed.