Asking questions about your data

How to phrase a question so the answer is the one you wanted. Specificity does most of the work, and the period does the rest.

How to phrase a question so the answer is the one you wanted. Specificity does most of the work, and the period does the rest.

Ask one

1. In Analytics, click the Home icon in the sidebar.
2. Enter the question, including the period and the grouping, then press Enter.
3. Read the report it produced, filters included.
4. Read the memories listed under the answer. Those are the saved notes analyst copilot used, and they tell you what it assumed.
5. Refine the question rather than the report.
6. Save it if you will need it again.

Name the period and the grouping

"How many tickets did we solve per group last month, excluding spam" gets you something usable. "How are we doing" gets you something.

Say which date you mean

Created or solved. This produces an answer that is right and not what you asked, because both are defensible readings of the same sentence.

Use the words Explore uses

"First reply time" rather than "how fast we answer". Where your internal vocabulary differs from the product's, translating it yourself removes a guess.

Sanity-check the size

Does the total look like your business? An answer an order of magnitude out is usually a missing filter or the wrong dataset, and it is obvious the moment you look for it.

Refine the question, not the output

If the answer is wrong, add the missing constraint and ask again. Correcting the generated report by hand leaves you with something whose definition nobody can reconstruct.

Save the good ones properly

A question that becomes a regular report should be rebuilt as a proper report, named for the question, with its filters visible. Generated reports are drafts, and drafts should not end up on a dashboard.

See also

• Checking a generated report
• What analyst copilot does
• Setting dates and date ranges