Building your first report

A worked example from empty to useful, with the two decisions that matter more than the clicking.

A worked example from empty to useful, with the two decisions that matter more than the clicking.

Build it

1. Create a report and choose the ticket dataset.
2. Add the metric: tickets solved.
3. Add an attribute: by month.
4. Add a filter: the last twelve months.
5. Choose a column chart.
6. Name it for the question it answers, and save.

Decision one: write the question first

Before opening Explore. "Is our volume growing, and by how much?" That sentence tells you the metric, the grouping and the period, and it prevents the drifting that produces charts nobody uses.

Decision two: what would change

If the answer is nothing, stop. Reports that inform no decision still cost the time to build and the space on a dashboard, and they crowd out the ones that matter.

Clone before building from scratch

Find the closest standard report, clone it, and adjust. Faster, and it shows you how a working report is put together.

Check it against something you know

Pick a month you remember and see whether the figure is plausible. A report that is wrong by a factor of ten looks perfectly convincing as a chart.

Name it as the question

"Tickets solved per month, last 12 months" tells somebody what they are looking at. "Report 4" means that in six months nobody dares delete it and nobody uses it.

Then stop adding to it

The urge to add one more attribute is strong and it is how a clear chart becomes an unreadable table. A second question deserves a second report.

See also

• How a report is put together
• Choosing the visualisation that makes your point
• Cloning a report