Two versions, one task. Results take about a minute.
PNG, JPG or WebP, up to 10 MB each. To test a whole flow such as a checkout, add up to 8 screens per version in order.
Phones get less patience and a 44 px tap-target rule.
800 simulated people in total.
The same inputs always give the same result. More people only smooths the percentages.
Screens are stored to show the result. Anyone with the link can open it.