Measure Anything Now, Decide What Counts Later
This release changes the order of the work. Until now, measuring something meant deciding upfront that it mattered, configuring it, and waiting for data to accumulate. Now your site can report everything it does, and you decide later what counts as a conversion. Define a goal today and its history comes back with it.
Capture events the way that suits your site
Your site can report any event by name, with optional properties and revenue, through the pagent snippet. Nothing has to be pre-configured: the label shows up on its own, ready to build goals on.
If you’d rather not write code, pagent can watch for browser events, clicks on a CSS selector, and pushes into your data layer, including gtag-style calls, with conditions on the payload. The tracking you already have becomes a pagent event.
To find out what your pages actually emit, the browser extension has a new Events view. Arm it, reload, and click through the page. It lists every event that fired, from dispatched events and data layer pushes to ad-platform pixels and network calls, grouped by name, with counts, the pages they fire on, the shape of their payload, and a flag on anything that looks like personal data. Define the ones you care about straight from the recording, and it verifies them against fresh occurrences without another reload. A new Events tab on the tracking page then shows everything pagent received: the source, the volume over the last 30 days, and when each was first and last seen.
Goals that apply to data you already have
A conversion goal is now defined over an event name and computed at the moment you look at results, not when you set it up. That means it applies retroactively to everything already collected. Spot a pattern this week and you can measure it against last month. Deleting a goal never deletes data, so you can try definitions freely.
Those definitions go well beyond “did it happen”. A goal can count once per session or once per visit, filter on the event payload, and require a per-visitor threshold, for example only visitors who did something at least three times. It can also measure magnitude: count occurrences, sum a number from the payload, or use revenue. Average order value and revenue per visitor are now tested metrics like any other, with credible intervals and a probability that a variation beats control. Revenue results are capped against outliers, so a single unusually large order can’t decide a test on its own, and the largest order values stay visible so you still see the whale.
A few more upgrades round out this release. Each goal can carry its own conversion window, counting only within a set number of hours after a visitor saw the test, ignoring the first stretch after exposure, or counting only the last stretch, all recomputed the moment you change it. Adding a click goal shows you how much traffic that click would have had before the goal existed, drawn from click history pagent already recorded going back several months, and a new Clicked Elements tab lists what visitors actually click most so you can create a goal straight from the list. Every goal has a detail page with one timeline of every test that ever targeted it, when each shipped and what lift it measured, plus a per-test ledger, and deliberately no grand total, because lifts from different tests don’t add up. And attribution is now a lens you can switch: set the model, scope, and lookback as a default for your site or as an override on a single test, or use “view as” to recompute every number on the page under a different model without changing anything stored, with every result naming the model behind it.