Skip to main content

Overview

How visits and goals are calculated​

Results are calculated on a unique visitor basis. This means each visitor is counted only once, no matter how many times they visit. Likewise, each goal is only counted once per unique visitor. This approach ensures that results aren’t skewed by highly active visitors who trigger the same goal multiple times or make repeated visits to the site.

Goal blocks​

Each goal in the A/B test is displayed in its own section. The primary goal appears first, followed by a quick-view graph of the secondary goals, which are then listed in full detail below. Each goal (for example, Purchase, Add to cart, or CTR checkout) represents a specific user action being measured. The primary goal reflects the main metric the test aims to improve, while secondary goals provide additional insights into related user behaviors. Within each goal section, all variations (such as Original and Variation 1) are shown with their corresponding performance data for easy comparison.

Variations​

Each variation displays the following metrics:

Variation​

Name and ID of variation.

Visitors​

Visitors show how many unique visitors that were assigned the variation.

Conversions​

Conversions show how many unique visitors that triggered the goal. In other words, even if a visitor triggers the goal multiple times it will still be counted as one conversion for that goal.

Conversion Rate​

Conversion Rate represents the estimated likelihood that a visitor completes the selected goal.

In a Bayesian test, this value reflects our current belief — based on the data — about how likely users are to convert for each variant, rather than a fixed point estimate.

Improvement​

Also called uplift. This shows how much better (or worse) a variation is performing compared to the Baseline. A positive number means it’s estimated to be doing better, while a negative number means it’s likely doing worse.

Probability to beat Baseline​

This is the chance that this variation is truly better than the Baseline. For example, a value of 95% means there’s a 95% chance this variation outperforms the Baseline.

Additional information​

Winning variation​

A variation is marked as a winner when the probability that the variant beats the control is at least 95% and the lower bound of the uplift’s credible interval is greater than +2%.

Why the numbers differs for each load​

Because these numbers are calculated using a Bayesian model, they update as new data comes in. Early on, the estimates are more uncertain and can move quite a bit from day to day. As more visitors and conversions are collected, the model becomes more confident — and the values stabilize.