Compare

Compare exists for one question: did the thing I did work? A number on its own cannot answer it. A number against a baseline can.

How to use it

1. Set the date range to the period you want to judge (the week of the launch, the month of the campaign).
2. Pick a baseline:
   • Previous period compares against the span immediately before. Best for “is this week better than last week”.
   • Year over year compares against the same dates last year. Best for anything seasonal, where last month is a bad baseline and last December is a good one.
3. Read the chart: the solid line is now, the dashed line is the baseline.

The metrics bar above the chart carries the deltas for the same comparison.

Reading it honestly

• Match the range to the change. Comparing a seven-day range that contains a launch against the seven days before it measures the launch. Comparing a 90-day range does not; the launch is a rounding error inside it.
• Watch the days of the week. Previous period on a 5-day range compares Monday-to-Friday against Wednesday-to-Sunday. Traffic that swings on weekends will look like a change that never happened. Use whole weeks, or year over year.
• One change at a time, or none of them are attributable. If you shipped a redesign and posted on social the same day, Compare tells you something changed and nothing about which.

Next: Reports.