Call Me On 0807 708 8529

A / B tests are a necessary element of any advertising campaign. Contextual, like other types of advertising, needs their application.

The main purpose of this research is to find out from the set of ads that which will provide the greatest CTR and conversion.

Features of split testing of contextual advertising:

  • The complexity in the selection of similar testing periods – the demand for a particular product / service at different time periods will be heterogeneous. For reliability of testing it is necessary to choose such time intervals that the level of demand was approximately equal for each tested advertisement. Often, a “chess” method of split testing is used for this.
  • The need to conduct research not in the scale of the ad group, but for each ad individually . This is necessary for a more accurate understanding of the block of advertisement that increased the effectiveness of advertising.
  • The long period of the test . This is due to the fact that the demand and preferences for many goods are very unstable (they can fluctuate depending on: weather, exchange rates, holidays and weekends, info-reasons, etc.). Based on this, A / B testing should be carried out until a stable “leader” is identified. Often this period is stretched for several weeks.
  • Inhomogeneous CA is a factor that increases the testing period. Proceeding from the fact that each segment of CA shows the greatest activity in different periods of time, it is necessary to increase the time interval for the study of all segments.
  • The difficulty in testing low-frequency queries . Testing low-frequency requests is perhaps the biggest problem of testing ads in contextual advertising. To collect relevant data, test ads must collect a certain number of clicks (and this is not 5, 10 or 20, but at least 100). What if there are only 20-30 impressions per month for certain low frequencies? Unfortunately, in this case, the test period is stretched for several months. Of course, you can limit yourself to a much smaller number of clicks for statistical data, but such statistics may not be true.

What can I test in contextual advertising in the search

The object under test is interpreted in 2-4 variants:

The element

Example

Title – it is desirable to use all the number of valid characters in it. However, the most successful heading, in our opinion, may turn out to be low-conversion. Therefore, you should come up with several variations of the title. The key phrase: “Buy a sofa inexpensively.”

Headline Option 1: “Buy a sofa inexpensively.”

Headline Option 2: “Buy an inexpensive sofa.”

The extended header is the same situation as the header. To select the most effective extended header is possible only by the method of split testing. The key phrase: “Buy a sofa inexpensively.”

Option extended title number 1: “Buy a sofa inexpensively – delivery in St. Petersburg.”

Option extended title number 2: “Buy a sofa inexpensively – round the clock delivery.”

Ad Text – in the ad text, we need to pack information that is brief but capacious will convey the advantages of the product / service. Identify the most conversion text ads will help A / B testing. The ad text must contain the benefits, advantages of the product / service or call to action. The key phrase: “Buy a sofa inexpensively.”

Variant of the text of the announcement number 1: “Large selection of sofas. Have time to buy at a bargain price. ”

Variant of the text of the announcement №2: “Free delivery in the Moscow region during the day.”

Specifying a price is a factor that can both attract and repel users. The need to specify the price, as well as the format in which it should be better indicated, should be identified by testing several options. The key phrase: “Buy a sofa inexpensively.”

Option number 1 with the price in the “header”: “Buy a sofa from 3000 rubles.”

Option number 2 with the price in the “ad text”: “A large selection of sofas from 3000 rubles. Delivery in Moscow for free. “

Quick links are an area where you can further outline the benefits of your product / service by adding additional “active” links. The key phrase: “Buy a sofa inexpensively.”

Variant of key phrases №1: «Prices | Shares | News | Delivery conditions».

Variant of key phrases №2: «Catalog of leather sofas | Catalog of suede sofas | Prices | Contacts».

Updating is another way to convey the benefits of the product / service to the user. The key difference between “refinements” and “quick links” is that the refinements do not have “active” links. The key phrase: “Buy a sofa inexpensively.”

Option number 1: “Delivery on time | Favorable offers | Working 24 hours | Buy on credit»

Variant of updates №2: “The best prices in Moscow | Purchase on installments | Free delivery | Daily actions”

What can I test in contextual advertising in the RSA

The object under test is interpreted in 2-4 variants:

The element

Example

The image is the part that attracts the most attention to the advertisement. Proceeding from this, the image should be maximally converted. To choose such an image is intuitive, but it is better to rely on statistical data and perform a split test. The key phrase: “Buy a sofa inexpensively.”

Variant of the image №1:

"

Variant of the image №2:

"

A header is a block that is the main trigger of an ad. To find the most suitable title is possible by the method of split testing. The key phrase: “Buy a sofa inexpensively.”

Headline Option 1: “Buy a sofa inexpensively.”

Headline Option 2: “Buy an inexpensive sofa.”

The ad text is a block that conveys the advantages of the product and prompts for purchase. The key phrase: “Buy a sofa inexpensively.” The ad text must contain the benefits of the product / service or call to action. The most conversion text is calculated by A / B testing.

Variant of the text of the announcement number 1: “Large selection of sofas. Have time to buy at a bargain price. ”

Variant of the text of the announcement №2: “Free delivery in the Moscow region during the day.”

Refinements – allow further stress (based on the fact that the number of characters in the text and the title of the ad is limited) the benefits of purchasing the product / service. Identify the most effective “refinements” with the help of research statistics on the displays of different variants of “refinements.” The key phrase: “Buy a sofa inexpensively.”

Option number 1: “Delivery on time | Favorable offers | Working 24 hours»

Option clarification number 2: “The best prices in Moscow | Purchase on installments | Free Shipping»

Methods of testing

There are two main methods of testing: simultaneous and in turn. Each method has its own pluses and minuses.

The method of the “chess player”

Means the creation of 2-3 ads, which are shown in turn. For example: 0-1 hour – the first variant of the announcement, 1-2 hours – the second variant of the announcement, 2-3 hours – again the first announcement, 3-4 hours – again the second announcement, etc. There are various options for “breaking down” the ads into time slots (by hours, by day of the week, etc.). A big plus of this method is that the external conditions (day of the week, time of day, etc.) are similar for both ads.

Or so:

"

Simultaneous launch of several advertisements

The essence of this method is that Yandex and Google show ads in turn, and after a while there remains only the most clickable ad (with a large CTR). However, this method has a serious drawback: the system rotates ads unevenly (at different times of the day, on different days), so the CTR of these ads is often formed under various external factors, and therefore the reliability of the test results is relative.

The main mistakes in the A / B testing of contextual advertising

  • Unreasoned ad units (all content of ads should be qualitatively elaborated).
  • Testing without preliminary analysis. To get high-quality A / V testing, you need to learn the statistics of impressions, Webview, competitors’ ads and other sources that will help analyze the effectiveness of existing ads.
  • Lack of detailed analytics during the test and its outcome. The results of testing should “flow” from the statistical data, and not be determined “by eye.”
  • Wrong time for testing. It should be understood that the demand and needs of users fluctuate depending on external factors, therefore, when planning a split test, periods should be avoided when the external factors are particularly strong (holidays, strong economic and political fluctuations, cataclysms, etc.).
  • Test multiple ad blocks at once. It is necessary to understand that for the correct analysis of the results of split testing, only one ad unit should be changed at a time (text only, only the title, only quick links, etc.), otherwise it is extremely difficult to identify which block affected the conversion dynamics, It is extremely difficult to correctly estimate the statistics obtained.

A few tips for conducting quality A / V testing

  • Always test the changes in only one block (only the ad text, only the title, etc.) in order to be able to analyze test statistics as efficiently as possible.
  • Do not run split testing in periods when external factors actively influence demand (holidays, economic and political fluctuations, change of seasons, etc.), because The obtained data will not be objective.
  • Always analyze the market, competitors, statistics of your own advertising campaigns before testing.
  • Test all the blocks in turn, trying to identify the most optimal and conversion.
  • Timing of testing should be sufficient so that each of the ads can collect 100-200 clicks. Analyze only a sufficient data set. Insufficient statistics are highly likely to be unrepresentative.
  • Carry out a thorough analysis of the data obtained, the result of which will be based solely on figures.
  • The results of the split-testing are relevant only for a particular product in a specific time period.
  • It is also important to analyze traffic for advertising phrases. If the phrase was more than 20 conversions, it makes sense to see the percentage of refusals on it in Yandex.Metrics. If it is close to 100%, then the displays for this phrase can be disabled. And in the future, or do not use it at all, or target another, more appropriate page.
  • If you want to test the operation of low-frequency queries, you can try here except to manually set the maximum bet for them. To get the largest number of transitions and generate statistics for the study. But this method does not always work.

Leave a Reply

Your email address will not be published. Required fields are marked *

Copy Protected by Chetan's WP-Copyprotect.