How changes are measured
When you measure a change, Lantern watches the pages and questions that change touched and compares them against their own prior trend and the movement of their category over the same days, not against the rest of your site.
Each change gets a verdict: moved when the numbers shifted beyond their normal range, watching while it is too early to tell, and no change when they stayed inside it. Watch windows vary by the kind of change, and the exact window is shown when you register one. Older changes keep the watch window stored when they were registered. Model releases and competitor launches that land inside a window are logged alongside the verdict, so a shift that lines up with an outside event is not mistaken for your change. When a watch window closes without enough traffic to decide, or the pattern it saw could have come from something else, the change is marked undecided.
Purchases appear in the Changes list by date. When the pixel sees an AI referred order, Lantern places it on the timeline beside whatever changes were live at the time. Sitting beside a change means the timing overlaps. It does not mean the change caused the purchase.
Where changes come from
Changes reach this list in three ways. You can measure one yourself with Measure a change. A page you publish from Content Studio registers as a change on its own. A fix you mark done in Opportunities registers as a change too. Lantern also lists changes it noticed on your site under Changes Lantern noticed, and measuring one of those adds it here.
Related
- Observed sales defines how purchase amounts are counted.
- Why counts differ explains why totals vary between views.