A control group in A/B testing represents your current product version — the "A" in A/B testing. Setting it up correctly ensures your experiment provides reliable insights about whether new changes actually improve the user experience.

Let's say you're testing a new checkout flow. Your current single-page checkout has a 25% conversion rate. To test a new 3-step checkout, you randomly split 10,000 daily visitors: 5,000 see the current version (control group A) and 5,000 see the new version (test group B). Random assignment ensures both groups have similar mixes of new and returning customers, mobile and desktop users, and peak and off-peak shoppers.

Traffic splits can vary based on risk. Testing a major change like a new pricing strategy? Use a 90/10 split (9,000 users see the current prices, 1,000 see new prices) to limit potential revenue impact. Testing a minor copy change? Stick with 50/50 to get results faster. Just don’t change these splits mid-test as they can skew your results.