Compare statistics without mixing time periods
Understand totals, selected periods, snapshots, and missing data before comparing dashboard numbers.
Tell a genuine reporting problem from a difference in scope or measurement.
On this page
Read the label before comparingCommon differencesMake a fair before-and-after comparisonIf a discrepancy remainsTechnical details and methodologyRead the label before comparing
Two valid numbers can differ because they count different events or cover different periods. A chat start, a message, a unique chat user, and a product click are not interchangeable. Read the small label alongside the headline, especially when a card mixes period activity with a saved total.
Common differences
Use these checks when the cockpit does not appear to match a detailed report.
| What you see | What to check |
|---|---|
| Chats differ from messages | One conversation can contain several messages. Compare chats with chats, not with messages. |
| A period chart differs from a total | Match the date range and distinguish saved overall statistics from the selected period. |
| The score has no daily line | A latest assessment is a snapshot. A trend requires comparable historical measurements. |
| A percentage changes sharply | Check the number of eligible observations. A small sample can move substantially after one new answer. |
| A dash appears | The measurement may be unavailable or incomplete. Do not replace it with zero. |
Make a fair before-and-after comparison
Use equally long periods, the same workspace, and the same event definition. Note major changes such as launching a new widget, changing questions, switching providers, or importing a different catalogue. These can change the population being measured even when the dashboard is working correctly.
- Write down the date and scope of the change.
- Wait until the next relevant assessment or usage period contains new data.
- Inspect a few individual records as well as the headline metric.
If a discrepancy remains
Capture both metric names, their periods, the module, and the expected relationship. A screenshot of the two labels is more useful than a statement that the numbers are wrong. Include no customer conversation content unless it is necessary and redacted.
Technical details and methodology
For the collection steps, eligibility rules, calculation units, and limits behind this workflow, use the implementation-reviewed technical guides below.
Still need help?
Tell us what you expected and what happened.