How advertising planners do AB testing

From data to large model application, on November 25th, Hangzhou Yuanchuang Conference was held to share development tips.

For more technical exchanges and job opportunities, please follow the WeChat official account of ByteDance Data Platform and reply [1] to enter the official communication group

In recent years, with the psychological needs of consumers gradually tending to the spiritual level, the rapid iteration of technology, and the complexity of the market environment gradually increasing, my country's advertising industry has gradually launched new advertising models. However, the cost of traffic is high and the cost of trial and error is high for enterprises. How to maximize the effectiveness of advertising while controlling costs has become the focus of advertising engineers.

The purpose of advertising is to maintain the continuous exposure of the company's products and help the company obtain more benefits. Therefore, advertising engineers need to control the cost of advertising while ensuring that the effectiveness of advertising is maximized. With the disappearance of Internet dividends, how to simultaneously control advertising costs and achieve expected advertising effects, so that advertisers feel "value for money", has become an increasingly difficult balance for advertising engineers.

AB experiment is a small-traffic random experiment. It has been widely used in business application scenarios such as product update iteration and marketing. In fact, AB experiment can also be widely used in the field of advertising.

So, how to apply AB testing to optimize advertising? Let’s talk about it in detail in this article.

The application of AB testing in advertising can be divided into five specific experiments: material split comparison experiment, placement crowd comparison experiment, H5 landing page optimization experiment, advertising efficiency measurement experiment and channel budget allocation experiment. This article will focus on these five experiments to introduce in detail the specific application of AB testing in this advertising scenario.

Material splitting and comparison experiment

In order to achieve the best marketing effect of advertising, advertising designers need to select the materials with the greatest traffic potential among the existing advertising materials. The material split comparison experiment can identify advertising materials with greater potential to become hits.

Therefore, advertising engineers can use AB testing to start material splitting and comparison experiments for different existing advertising materials, and quickly determine which type of material has a tonality that is more in line with the selling point of the product promotion. Select the advertising materials with the highest conversion rate based on experimental data indicators to improve advertising efficiency.

Target population comparison experiment

Different groups of people have different preferences for types of advertising materials. Accurately pushing advertising materials that match their preferences to groups of people can effectively improve the conversion rate of advertising.

Therefore, in order to analyze which group of people is more sensitive to which type of material and identify the group of people who are better matched with the product and material, after conducting the material split comparison experiment, the advertising placement engineer also needs to start the placement group comparison experiment based on the audience group. After classifying the possible audiences, place different advertising creatives for different groups of people to set up multiple experimental groups and start AB experiments. Based on the click-through rate, conversion rate and other data indicators obtained from the experiment, select the people who match the creative for delivery.

H5 landing page optimization experiment

The landing page is the first step to receive advertising traffic, is the key to user conversion, and plays an important role in the application scenarios of advertising. Therefore, advertising designers need to conduct H5 landing page optimization experiments. Starting an H5 landing page optimization experiment requires three steps: determining test content, traffic allocation, and agile adjustments to optimize iterations.

First, you need to determine the content of the test . Swapping button positions, changing colors, adjusting order, etc. will all have an impact on the increase in conversion rate. Advertising engineers need to start AB experiments for the above content respectively, and select the best plan based on data indicators such as background click rate and conversion rate.

After determining the test content, we need to allocate traffic between the A/B versions . There are generally two goals when designing tests: getting experimental conclusions as quickly as possible and minimizing the impact on the user experience. Therefore, it is necessary to make trade-offs in traffic allocation, and choose to evenly distribute traffic experiments, small traffic experiments, or large traffic experiments according to the specific content of the experiment.

In the AB testing process, agile adjustment and optimization iteration are the most important. Advertising engineers need to continuously adjust and iterate the landing page based on experimental data to adapt to the changing needs of users and improve conversion rates.

Advertising effectiveness measurement experiment

The advertising efficiency measurement experiment is a combination of crowd diversion and questionnaire delivery to evaluate the effectiveness of product promotion. After the advertisement is placed, the advertising placement engineer needs to make placement adjustments based on the effect of the advertisement placement, so the advertising effectiveness measurement experiment can be started.

Through online control experiments, the groups of people who invest in advertising and those who do not invest in advertising are constructed into a virtual experiment. Compare and judge the conversion differences, behavioral differences and user cognition differences between the two groups of people within a period of time after the advertisement is placed, and then scientifically measure the advertising value and verify the impact of the advertising placement on the audience's subsequent conversion behavior. Then judge the effectiveness of advertising creatives. Based on the results of this experiment, the advertising team will then carry out follow-up actions such as redistribution of advertising traffic.

Channel budget allocation experiment

Advertisers have a limited budget for advertising. To achieve the best advertising results, they need to allocate the budget reasonably and make sure "every penny is spent wisely" as much as possible.

Therefore, advertising designers need to apply AB testing to conduct channel budget allocation experiments. Start AB experiments for advertising channels and advertising pages. Carry out delivery effect analysis based on the background click-through rate, conversion rate and other data indicators obtained from the experiment. Compare and select delivery channels and pages with better results, and then allocate budget appropriately.

AB testing has a wide range of applications in the advertising industry, covering the entire process from the production of advertising materials to the implementation of advertising. In addition to the above-mentioned basic AB experiments, Volcano Engine DataTester also provides cross-channel advertising capabilities, integrated website building platforms, opening up crowd data, opening up front and rear link data, and intelligent experiment optimization for enterprise advertising scenarios. . With the help of AB testing, advertising designers can achieve a balance between delivery costs and effects, and optimize advertising delivery.

Volcano Engine DataTester originated from ByteDance’s long-term accumulation. As an AB testing platform that helps enterprises make scientific decisions, DataTester currently serves hundreds of companies including Midea, Get, BSH, Kaishu Storytelling, etc., providing business Provide scientific decision-making basis for user growth, conversion, product iteration, operational activities and other aspects, and empower various industries with mature "data-driven growth" experience.

Click to jump to DataTester to learn more