A difference between two averages can be real or just the luck of which customers happened to arrive. t.test(order_value ~ variant, data = checkout_orders) asks how surprising the observed difference would be if the two groups had the same true average.
The result is a list. $p.value is that surprise: below 0.05 is the usual line for calling a difference statistically significant. $conf.int is a range for the true difference (first group minus second) that is consistent with the data; if it includes 0, no difference is still plausible. $estimate holds the two averages.
R's default is Welch's test, which does not assume both groups vary by the same amount. Keep it unless you have a reason not to.
When the same items are measured twice, such as each class before and after a schedule change, compare each item with itself: t.test(after, before, paired = TRUE). Treating the two columns as separate groups throws away that pairing and usually hides real changes.
A two-group test, its p-value and interval, and a paired test.