Measure conversions

You define what counts as success on your site by marking one or more signals as conversion goals.

A signal already describes something a visitor did: submitted a form, watched a video to the end, reached the checkout, or spent a while on a pricing page. Marking it as a goal makes it a metric: it becomes selectable in the composition editor's measurement criteria, appears in the dashboard's conversion rate widget, and is something you can ask Scout about.

Prerequisites

  • The Manage insights permission, which team admins have by default.
  • A signal that fires when the thing you want to measure happens. Create one first if you do not have it.

  1. In Uniform, open your project and navigate to Optimization > Signals.

  2. Open the signal you want to use.

  3. In the Trigger Conversion panel, select Once.

    note

    Never is the default and means the signal is not offered as a goal in Insights. Once counts a conversion the first time the signal fires for a visitor within the signal's configured duration, so one visitor converting repeatedly counts once rather than inflating the rate.

  4. Save the signal.

  5. Publish the manifest.

Publishing is required

Conversion tracking runs from the published manifest in your front end. Until you publish, and depending on your setup possibly rebuild and redeploy, no conversion events will be recorded.

Once a goal is configured and data starts arriving:

  • The conversion rate widget appears on the dashboard, showing the overall rate, the share influenced by optimization, and the top five goals for the period.
  • The Measurement criteria menu in the Insights tab gains an All goals entry and one entry per goal.
  • The dashboard's active tests and active personalizations tables can calculate uplift.
  • Scout can answer questions scoped to a named goal.
  • Fewer goals read better than many. The dashboard charts the top five, and a test measured against a goal that fires rarely will take a very long time to reach significance.
  • A goal should be downstream of the thing you are testing. Measuring a homepage hero test against a goal that only fires deep in a checkout flow will mostly measure the funnel between them.
  • Goals can come from your existing data layer. Signals can be triggered from events you already emit, so you do not have to reimplement your analytics instrumentation to get conversion data into Insights.

Conversion counts, not traffic, are usually what limits how quickly a test resolves. Insights requires each variation to have at least 200 sessions and enough expected conversions among them before it will report a verdict, precisely because a rate computed from a handful of conversions is noise. See how significance works.

Last modified
Next
Ask Scout about Insights