A/B testing in traffic arbitrage makes it possible to make decisions based on actual data rather than subjective assumptions. Instead of choosing a creative, landing page, or audience based on the principle that “this version looks better,” a webmaster launches several versions of a campaign setup and compares their performance under comparable conditions. This approach is especially important in situations where even a small change in CTR, CR, or acquisition cost can significantly affect the overall profitability of a campaign.

The essence of A/B testing is to compare two or more versions of a single element within an advertising setup. For example, you can use two different creatives, two headlines, different landing page versions, or several approaches to a call to action. The key objective, however, is not simply to find the version that generates more clicks, but to determine which change actually improves campaign economics. Therefore, A/B testing in traffic arbitrage should take into account not only CTR, but also CR, CPA, approval rate, EPC, and overall ROI.

Why You Should Not Change All Elements at Once

One of the most common testing mistakes is changing several components of a campaign setup simultaneously. A webmaster may change the creative, ad headline, audience, landing page, and bid, and then compare the result with the previous campaign. Even if performance improves, it becomes almost impossible to determine what caused the change.

Imagine that CR increases from 4% to 6% after making several changes. If both the advertisement and landing page were changed at the same time, there is no way to know which factor affected the result. The new creative may have attracted a more interested audience, while the new landing page may have actually reduced conversion. In another campaign, the effect could be completely different.

Therefore, the classic approach is to test one main element while keeping the other conditions as comparable as possible. This makes it possible to connect a change in performance with a specific hypothesis and use the resulting insight for further optimization.

What Can Be Tested in an Advertising Campaign

The number of potential hypotheses depends on the traffic source, vertical, and campaign setup. In practice, almost any element that can influence user behavior can be tested: advertising creatives, headlines, copy, images, ad formats, CTAs, landing pages, forms, audiences, GEOs, and even individual stages of the funnel.

A/B tests are particularly common for advertising creatives. You can compare different visual concepts, main message formulations, text length, or call-to-action variations. At the same time, it is important not to turn testing into a chaotic process of trying random options. Every new creative should be designed to test a specific hypothesis.

For example, the objective may be to determine whether a more specific advertising message affects audience quality. In this case, two versions are compared that differ specifically in the wording of the offer, while the other campaign parameters are kept as consistent as possible.

Testing Headlines and Advertising Messages

The headline creates the user's first impression of an advertisement and directly affects the likelihood of a click. However, an overly aggressive or exaggerated headline can increase CTR without producing a corresponding increase in conversions.

Therefore, when testing headlines, it is necessary to evaluate subsequent user behavior. The version that generates more clicks will not necessarily be more effective in terms of CPA. If the additional clicks mainly come from users who do not complete the target action, advertising costs may increase without generating additional revenue.

It is much more informative to analyze the entire chain: impressions → clicks → visits → conversions → approved leads → revenue. This approach helps determine whether a new headline attracts genuinely relevant users or simply increases the number of clicks.

A/B Testing Advertising Creatives

Creatives are one of the most obvious elements to test. Even with the same audience, different visual and textual approaches can significantly change user response.

When comparing creatives, it is useful to determine the primary test metric in advance. If the objective is to increase click-through rate, CTR will be the main indicator. If the goal is to reduce the cost of a target action, the analysis needs to go deeper and include CR and CPA. When working with a CPA offer, it is also important to consider the quality of the leads generated and their subsequent approval.

This is particularly important for campaigns where several stages exist between the click and the payout. A creative may demonstrate excellent CTR and low CPC but attract an audience that does not meet the offer requirements. As a result, once approval rate is taken into account, that version may prove to be less effective.

How to Test a Landing Page

A landing page affects conversion after the user has already clicked through from the advertisement. Therefore, with stable traffic volume and quality, A/B testing a landing page makes it possible to determine which changes help increase the number of target actions.

You can compare page structure, headlines, CTA placement, lead forms, the number of fields, information hierarchy, and other elements. However, the same principle applies here: it is preferable to change one significant element at a time.

For example, if version B simultaneously changes the headline, image, form, and CTA, and CR increases from 5% to 7%, the result is difficult to interpret. For the next test, it is better to preserve the successful foundation and test one new hypothesis. Gradually, this process makes it possible to build a landing page based on actual performance data.

Testing CTAs and Forms

A call to action is often treated as a minor detail, although its wording and placement can influence user behavior. Different audiences can respond differently to direct calls to action, informational wording, or messages that emphasize a benefit or the next step.

Forms can also be tested step by step. For example, you can compare a shorter version with a more detailed one or change the order of the fields. However, reducing the form simply to increase the number of applications is not always the right approach. If the advertiser requires specific information to process a lead, removing those fields may result in a higher rejection rate.

Therefore, the final metric should not be limited to form CR. The quality of the result after the lead is transferred to the advertiser should also be taken into account.

Can Different GEOs Be Tested?

The A/B approach can be applied not only to elements within a single advertising campaign but also to geographic markets. However, comparing GEOs requires particular caution because different countries can vary significantly in traffic costs, purchasing power, competition, user behavior, and the conditions of a specific offer.

If several GEOs are being tested, CPC, CR, CPA, approval rate, EPC, and ROI should be evaluated separately for each market. Simply comparing the number of leads will not provide an objective picture.

For example, one country may generate more applications at a higher acquisition cost, while another may produce fewer leads but have a significantly higher approval rate. In such a situation, the decision about future budget allocation should be based on the complete economics of each GEO rather than the number of initial conversions.

How to Determine a Sufficient Amount of Data

An A/B test is meaningless if conclusions are drawn after only a few clicks or one or two conversions. With small sample sizes, statistics can fluctuate significantly, and random changes in user behavior can create the illusion that one version is outperforming another.

There is no universal number of clicks or conversions that is equally suitable for every campaign. The required sample size depends on the baseline conversion rate, the size of the expected difference between variants, traffic costs, and other experimental conditions.

Therefore, before launching a test, it is useful to determine what level of change would be meaningful for the campaign's economics. If version B is expected to reduce CPA by only a few percent, significantly more data will be required than when testing a hypothesis involving a major change in conversion rate.

The main rule is not to stop a test simply because one version has temporarily produced slightly better results. The smaller the sample, the higher the probability that the observed difference is caused by random variation.

Which Metrics Should Be Used for Comparison?

The choice of the primary metric depends on the objective of the experiment. For a creative, CTR may be the initial benchmark; for the cost of a click, CPC; for a landing page, CR; and for the overall campaign economics, CPA, EPC, or ROI.

At the same time, no single metric should be considered in isolation. For example, a decrease in CPC may look positive, but if CR falls at the same time, the final CPA may increase. Likewise, an increase in CR may be accompanied by lower lead quality and a decline in approval rate.

Therefore, it is more useful to view the metrics as a sequence. CTR shows the audience's response to the advertisement, CPC helps evaluate the cost of acquiring a click, CR shows the ability of traffic to turn into a target action, CPA reflects the cost of a conversion, approval rate indicates lead quality, EPC measures traffic profitability, and ROI reflects the overall profitability of advertising spend.

Why a High CTR Does Not Always Mean a Successful Test

Suppose version A achieves a CTR of 2.5%, while version B reaches 4%. At first glance, the second creative appears more attractive. However, if users from version B convert at half the rate after clicking, the final cost of the target action may be higher.

This is one of the reasons why traffic arbitrage campaigns cannot be optimized solely for click-through rate. The purpose of an advertisement is not to generate as many clicks as possible, but to attract users who are sufficiently likely to complete the target action.

As a result, the best creative in terms of CTR and the best creative in terms of profit may be different versions. This is why testing should take into account both intermediate and final performance indicators.

How to Record Test Results

Every experiment should ideally be documented. It is enough to record the original version, the hypothesis being tested, the launch period, the amount of traffic, and the main results. This helps avoid repeating experiments that have already been conducted and gradually builds a knowledge base for specific traffic sources and verticals.

It is particularly useful to record not only the winning version but also the reason why a hypothesis was considered ineffective. After several months, accumulated data may reveal patterns that cannot be identified within a single campaign.

At the same time, external conditions must be taken into account. A test result obtained in one GEO or during a specific period will not necessarily be identical when traffic costs, audience composition, or offer conditions change.

How to Move from Testing to Scaling

Once a stable result has been obtained, the next question is how to apply it at a larger scale. It is important not to confuse a successful test with a guarantee that the same performance will remain unchanged during scaling.

Increasing the budget changes the size of the available audience, traffic costs, and sometimes the composition of users. Therefore, a version that performed well at a small scale should be scaled gradually while the key metrics continue to be monitored.

If CPA, approval rate, and ROI remain stable within acceptable limits after increasing spending, the hypothesis receives additional confirmation. If campaign economics begin to deteriorate, it is necessary to determine whether this is caused by audience saturation, rising traffic costs, or a decline in the quality of newly acquired segments.

A/B Testing as an Ongoing Process

In traffic arbitrage, it is impossible to find the perfect creative or universal landing page once and never change anything again. Traffic sources update their algorithms, audience behavior changes, competition affects advertising costs, and offer conditions may change. Therefore, testing becomes a permanent part of managing advertising campaigns.

13PARTNERS works with various traffic sources, GEOs, and monetization models, making a systematic approach to experimentation particularly important when comparing different campaign setups. The use of statistics, tracking, APIs, and postback integrations helps match conversions with specific traffic sources and segments more quickly.

The main value of A/B testing is not finding a random “winner,” but systematically accumulating data. Every tested hypothesis should provide insight into what specifically affects the cost and quality of conversions. When changes are tested sequentially, sufficient statistical volume is taken into account, and the analysis focuses not only on clicks but also on the final economics, optimization becomes a controlled process rather than a series of random experiments.