Second category e-commerce丨Practical information! Teach you how to do product launch testing!

In the previous issues, we talked about the production of delivery materials, helping everyone solve the difficulties in the production of copywriting, materials, and landing pages. Now that we have the materials, the next step is to place the advertisement. Today, I will analyze what is needed for delivery testing. Without further ado, let’s get started.

Many advertisers start advertising as soon as they have prepared the relevant landing pages. If the planned effect is not good, they will immediately change the audience package, materials, display methods, etc. However, we just cannot find the root cause of the problem. No matter how good the product and materials are, we cannot achieve the expected results. Because you are missing a key step - launching a test idea.

The launch test is not "trial and error" or "pitfalls", but to do variable testing with a strong purpose, so that errors can be found as soon as possible when there are loopholes in the plan.

Testing is essential in the early stages of launch. However, you cannot test blindly. You must test with a plan and ideas. If you don’t use your strength on the cutting edge, it will be useless. The correct way to launch is generally divided into three steps:

Ⅰ. Make two or more new a/b plans through imitation and creation

Ⅱ. Test and optimize by controlling variables

III. Finally, test out the bill that is most suitable for you and has the most prospects and launch it in large quantities.

So what is A/B testing?

A/B testing is a controlled experiment, that is, for the same conversion goal (for example, we want to see which landing page is more effective in submitting a form), we develop two or more plans (landing pages or main images). Under the premise of keeping other variables unchanged, we allow visitor groups with the same composition and characteristics to randomly visit these landing page versions. Based on the user data feedback from each group and combined with statistical tools and methods, we can screen out the version that better meets the effect requirements.

So, the key points to performing an effective A/B test are:

• N solutions are tested in parallel;
• There is only one variable among N solutions, eliminating the interference of other factors;
• Use a certain standard to judge the results and select the best solution.

Testing in different time periods means that other related variables will be inconsistent, such as the visitor's own visit cycle, changes in the content environment in the media, the impact of advertising from peer competitors, etc. These factors will interfere with the reliability of the test results.

Why do we need to control variables in testing?

It all comes down to the validity of the test. We all know that there are too many factors that affect conversion. Factors such as competition, the quality of traffic itself, and the consumer purchasing cycle may affect conversion. Especially in the context of information flow advertising, we cannot predict what kind of psychological arousal the information content that consumers see before seeing the advertisement will bring to consumers (which can also be understood as traffic shaping actively initiated by the platform). These will affect the advertising conversion effect.

After the test data comes out, we can optimize it based on the data and make our plan as perfect as possible.

Data analysis can refer to the classic four-quadrant rule of conversion evaluation:

In this four-quadrant rule, the horizontal axis represents our advertising expenditure, and the vertical axis represents the conversion volume generated by the advertisement. Let's look at them one by one:

The first quadrant: has three attribute labels, namely high conversion, high consumption, and high cost. For this type of data, there are many factors that need to be optimized, with the focus on bidding, click-through rate/conversion rate, and the accuracy of copywriting creativity. If the conversion rate is high, the click volume is high, the consumption is high, and the cost is high, you can try to reduce the cost and adjust the bid.

The second quadrant: This part contains the best advertising data, with high conversion and low cost. It can be continuously observed. Of course, customers with large budgets can also expand the volume appropriately.

The third quadrant: This part obviously means that the advertisement failed to be delivered and the exposure level is too small, so we need to give priority to the display volume when optimizing. If the conversion rate is low, the consumption is low, the cost is low, the click volume is low, and the exposure is low, you can first try to increase the bid and optimize the creative to increase the exposure.

The fourth quadrant: For the last part of the data, the conversion volume is low and the conversion cost is too high. Our priority is to optimize the delivery plan (including creativity/landing page/targeting, etc.) and the accuracy of the traffic. If the conversion rate is low, the click volume is low, the consumption is high, and the cost is high, you can try to optimize the creativity and targeting.

Next, we will briefly explain it through a case:

Plan A

Plan B

In this case, we can see two plans, A and B.

The characters and texts in the materials of Plans A and B remain unchanged, which are constants.

The background images of the material are one with a gray background and one with a gradient background, which is the variable.

Plan A has 11 clicks for every 1020 exposures, while Plan B has 45 clicks for every 5005 exposures. The click-through rate of Plan A is significantly higher than that of Plan B.

Then we come to the conclusion: the picture material of this warm clothing with a gray background or a simple background color is better than the gradient color effect.

Summarize

Testing a product requires consideration of: budget, cost, pricing, demographic targeting packages, platform, the program itself, and more. The test results for each product are different, so we cannot limit you to exact numbers. Therefore, we can only tell you the methods and ideas. There is still a long way to go, and I hope this will be helpful to everyone.

Author: Second-class e-commerce operation

Source: Second-class e-commerce operations