Analytics
From metric changes to grounded explanations
Agents compute metrics in code, break down changes and answer definition questions. Analysts confirm conclusions and every number that goes out.

Where work stalls
- 01
The same metric shows different values in different reports, so meetings start by reconciling definitions.
- 02
Ad hoc data requests take up most of the week, and deeper analysis keeps slipping.
- 03
Dashboards show that a metric dropped, but explaining why still means breaking it down one dimension at a time.
Work agents take on
01 / 04
Weekly review
Mondays at 7:30 a.m.
Update the weekly business review from this week's exports in WBR data. Give each metric's value, week-over-week change and gap to target, add a likely cause from the launch calendar where a change crosses the threshold, mark anything unconfirmed as a hypothesis and calculate every number in code
- Schedule · Mondays 7:30 a.m.
- Knowledge Base · WBR data / Week 46
- Launch and campaign calendar.xlsx
- Sandbox · code
- Check that every day is present and row counts and fields look normal
- Compute values, week-over-week change, target gaps and 8-week trends in code
- Explain metrics that crossed the threshold, marked supported or hypothesis
- Recheck the numbers and rewrite or cut the hypothesesAnalyst Nora Fields
- Send the review to business owners and save it to the library
Every step is in the run trace

Where people decide
Agents prepare the work. Key decisions stay with people.
- 01
Analysts confirm metric definitions and every number that leaves the team
- 02
Before a cause becomes a conclusion, an analyst judges whether the evidence holds
- 03
Product leads decide whether to roll out an experiment, based on the readout











