Fenwick Market tested a new checkout: half the sessions saw A, half saw B. Each session converted or not, so each group has a conversion rate. The two-proportion z-test asks whether the rates differ by more than chance.
In statsmodels: proportions_ztest(counts, nobs) with the conversions and sessions of each group, in the same order. alternative='larger' tests whether the first group's rate is larger.
Report three things: the two rates, the lift (B's rate divided by A's, minus 1, a relative change) or the absolute difference in points, and a confidence interval for the difference: confint_proportions_2indep(count_b, n_b, count_a, n_a) gives B minus A.
A t-test on 0/1 data gives nearly the same p-value for large samples, but the z-test is the standard tool and handles small rates properly.