What does the cookie banner a/b testing tool compare?
The CookieConsentHQ cookie banner a/b testing tool compares variant A, your current banner, with variant B, a version containing permitted design or text changes. It shows each variant's accept rate and, after enough decisions, the probability that B beats A.
The tool is available from Growth upward, including Business and Agency. The controls cover a second banner variant, the traffic split and displayed comparisons. If you want guidance on setting up a fair test and interpreting changes without dark patterns, read the separate cookie banner A/B testing guide. The guide addresses testing decisions, while this tool offers a defined set of banner changes and results.
A and B refer to the two versions being compared, with A remaining the current banner. The result is a comparison of recorded acceptance behavior for those variants. Do not read the displayed accept rate as a guarantee of the same rate on every future visit or on a different configuration.
What can I change in variant B?
Variant B can change the banner layout, button colour, title and Accept all button text. Layout options are box, bar and popup, and Reject remains available in both variants.
Those are the available editing boundaries for this test. A different layout lets you compare the listed presentation formats, while a colour or text change lets you examine the associated accept rates. Reject is not one of the fields a variant can change, so both versions always offer it. Keep that constraint in mind before you choose what to test.
If your planned experiment needs changes outside those four fields, the listed variant controls will not cover it. The tool does not offer separate variant edits to regional targeting or tracker-blocking behavior. For settings tied to visitor location, see geo-targeting; that is a separate product feature with its own plan requirements.
How do I choose the traffic split?
The tool lets you set the share of visitors who see variant B from 10% to 90%, in steps of 10 percentage points. The rest see variant A, your current banner.
This range gives you defined allocation choices in the product. For example, 10%, 20% and 30% are within the supported increments, as are 70%, 80% and 90%. A small share for B limits how many visitors see an untested design while the comparison runs.
The traffic split and the results threshold are separate settings and conditions. A split determines the allocation, while the probability display waits for the required number of decisions in each variant. When planning a test, keep those two facts distinct instead of treating a larger share of traffic as an automatic guarantee that the threshold has been reached.
When can I see which variant is ahead?
The results show an accept rate for each variant, and the probability that B beats A appears once both variants have at least 30 decisions. Before each variant reaches that threshold, the probability figure is not shown.
This is a display rule for a specific metric, not a declaration that a test has a guaranteed outcome after 30 decisions. A probability shown in the interface should be read alongside both variants' accept rates and the changes you made to B. The 30-decision threshold is the condition for showing the figure, not a universal sample size for declaring a winner.
For practical testing choices, the testing guide covers how to plan and evaluate a fair experiment. The tool reports the specified accept rates and probability; interpretation remains part of planning your test. The pricing page shows which plans include A/B testing if you are deciding whether to upgrade from Starter.
Can I trace a stored consent choice to its variant?
Yes, each consent record stores which variant the visitor saw. That variant information is saved alongside the other fields included in the cookie consent log.
The log includes a consent ID, UTC timestamp, action and selected categories, along with banner configuration version, location details and other recorded context. Your site can read a visitor's consent ID with CCHQ.getConsent().id, and you can open that record by ID in the dashboard. The variant field helps you identify the banner version associated with the recorded choice, rather than relying on the aggregate accept rates alone.
Records are also exportable as CSV, subject to plan retention and logging limits. Growth and higher plans keep records for five years, so the available log duration is the same across plans that include the A/B testing tool. The stored variant identifies what the visitor saw; the tool's results remain the place to compare accept rates and the probability that B beats A.