App's first launch of word selection routines-a strategy you will learn after reading it!

App's first launch of word selection routines-a strategy you will learn after reading it!
The overall process of word selection First, let me introduce the general process of word selection. The general idea is to "roughly select" first, then "finely select", and gradually narrow the scope of word selection. As shown in Figure 1:  Figure 1 The overall process of word selection The first step is to determine the “effective keywords”, which mainly involves basic popularity filtering. Because words with a popularity of 4605 are searched once a day on average, when choosing words, we generally choose words with a popularity greater than or equal to 4605. As for other words with lower popularity, most of them are long-tail search words covered by keywords and do not need to be considered deliberately. There are approximately 46,000 words in the appstore with a value greater than 4,605 ​​(Note: after mid-July, the number of words dropped to less than 30,000), and approximately 60,000 words with a value equal to 4,605. These words are generally referred to as “effective keywords”. The second step is rough selection. Mainly based on the relevance of the words, 100 to 200 words are extracted from the "effective keywords". The rough selection process is generally also called "keyword expansion". The commonly used methods are shown in Figure 2. The relevance of the obtained keywords decreases in turn, but the scalability increases in turn: 

Figure 2 Commonly used word expansion methods 1) Extract from the introduction. That is to analyze the keywords in the App introduction and find the words with popularity >= 4605. Generally, it is necessary to segment the introduction first, and then extract the words with higher popularity. 2) Category words. Just look at the keywords under a category, such as "weather". The category of a keyword is determined by the first app main category hit by its search. If your app is very popular, just choose the most popular words in the category. Some apps that are refreshed by machine also select all the popular words in the category and then refresh them. 

Figure 3 Category hot words in appbk 3) Competitive product word selection. This is relatively easy to understand. Just look at what keywords your competitors are using. 

Figure 4 Competitive product word selection in appbk 4) Choose words based on their meaning. Those who are familiar with Baidu's advertising platform may have used similar tools . This type of word selection method generally involves first filling in a batch of seed words, and then expanding them through different methods. There are three common types of extension methods: 
  • Words used by related apps. This method is one of the most commonly used recommendation methods, similar to buying a book on Amazon, which will recommend to you, "Customers who purchased this product also purchased ***." The same principle applies to the words used by related apps, that is, "apps that use these seed words also use the keyword ***." The specific principle is shown in the figure below:
 Figure 5 Words used by related apps 
  • Contains words. That is, the words that contain the seed word. For example, if the seed word is "飞车", then in this way, words such as "天天飞车" can be expanded.
  • Related words. A typical example is synonyms, such as the synonym "moving" which is "relocating". There are also some broader extensions, such as "beauty" can be expanded to "handsome guy", and "Chinese" can be expanded to "mathematics". There is a certain correlation between these words. For example, Apple’s search ads specifically introduce this type of “related word” expansion. For example, “photo filters” can be expanded to include words such as “picture editor” and “light right”.
 

Figure 6 appbk’s word selection service The third step is careful selection. The main thing is to select the final keywords of 100 characters from the roughly selected one or two hundred words. The selection range is much smaller. With detailed selection, you can define specific rules to choose words based on your understanding of the product. Generally speaking, the first words selected can be based on the popularity range. When updating the selected words, it is generally necessary to estimate the ranking of the new words and predict the volume. The process of "rough selection" is basically the same for both first release and updated word selection, the main difference is the "fine selection" strategy, so this article will mainly introduce the "fine selection" in first release word selection, and updated word selection will be put in the next article.  The strategy of “fine selection” for first release words: word selection in different regions When an app is first released, there is no online information. Therefore, the main goal of word selection should be to test the word selection and positioning of the app. If it is a niche app, it is best to choose some low-popular words, but if you are recommended by WeChat, then it is better to use popular words. But before going online, these are all unknown, so it is best to try as many words of different popularity as possible to achieve the purpose of "testing the waters". Because you can only choose 100 characters, which is about 30 words, the best strategy is to select words according to the popularity distribution of keywords. For example, there are only about 3,000 words with a popularity greater than 5,000, and there are about 43,000 words between 4,605 ​​and 5,000. Therefore, when selecting words, you can select 10 words for testing if the number of words is above 5,000, and you can select more words if the number of words is below 5,000. Try to keep the number of words consistent with the popularity distribution, so that it will be easier to find the word selection location. This method of selecting words based on heat distribution can be briefly described as "partitioned word selection". When we select words for the first release at Appbk, we generally use the following partitioning strategy, which is for reference only: A total of 130 characters were selected and divided into 6 groups from G0 to G5. The specific groups are shown in the following table: Table 1: Partition word selection

 When choosing words, follow these steps: 1. Select words from the high-heat interval to the low-heat interval (G5 to G0) in sequence. The number of characters selected in each interval should be greater than or equal to the preset number of characters, and they can be uniformly deleted later. 2. In each heat range, words are selected mainly according to competitiveness, from low to high, and selected from the candidate words in turn until the preset number of characters is exceeded. If there are not enough words to choose from within the interval, just select all the candidate words. The competitiveness of our appbk is mainly calculated based on the number of App reviews . 3. After selecting the 130-character word, adjust the ratio of "competitor words" and "industry words" according to the characteristics of the product. If no adjustments are made, the ratio of "competitor words" and "industry words" in the selected words should be roughly 15%:75%, which is the ratio of the overall market. Generally, the ratio of "industry words" can be appropriately increased. 4. Depending on the word formation, delete appropriately to ensure 100 characters. Combined with appbk or related tools, you can easily select the first words according to the above process. What? I still think it’s too troublesome@#￥@#￥@#. Because the rules of this process are relatively clear, most of the work can actually be done by machines. According to this rule, we are developing a "one-click word selection" system. Currently, it is mainly used by our internal ASO optimizers to improve it. This system will be opened as soon as possible in the future:) 

Figure 7 The “one-click word selection” function under development 

<<:  Those who only look at the CTR of information flow ads are all hooligans! Attached effect case

>>:  3 ways to write “resonant copy”

Recommend

What are Baidu bidding creative punctuation marks?

Baidu Creative's sentence break is used to de...

The sales of these 6 Korean SUVs have been pushed into a desperate situation.

In the automobile market, SUV is "undoubtedl...

How to improve mobile app performance with a powerful database

You may be aware that there are seemingly endless...

How much does it cost to develop a shopping mall mini program in Liaocheng?

There is no fixed price for the development of th...

On the second day of the Chinese New Year, I want to be the leader of the praise group! Learning to praise is really important

There is a saying in psychology: " People al...

Visual immersion is not enough, VR still needs multi-dimensional sensory stimulation

Recently, the concept of "virtual reality&quo...

Online shopping scam prevention tips summarized by experienced poor people can save you money!

One minute with the doctor, the postures keep cha...

Harvard Investment Practice Course: Learn these 10 investment strategies and earn an extra 3 months salary every year

Course Description: Suitable for the crowd: *Ever...

Raw fish, wolf fin, and "Liao" raw fish soup: This fish shares a "circle of friends" with dinosaurs

In the primary school Chinese textbook, there is ...

Madhouse TikTok Dou+ Nuggets Training Camp (Second Session)

: : : : : : : : : : : : : : : : : : : : : : : : : ...

How to design an e-commerce new product channel?

Nowadays, with the rapid development of the Inter...

The whole story of the suspension of LeEco's supercar Faraday factory: illusion, bubble, scam and the driving force behind it

North Las Vegas, Nevada. Although it has a North ...

Tesla to use rare earth permanent magnets to drive electricity in preparation for entering China

Tesla purchased neodymium iron boron magnets for ...

The new mirror for the young lady? 360 Mobile Phone N6 Pro has a new way to play with the full screen

With the advent of full-screen phones, the strate...

After 7 years, Logitech launches the new trackball mouse MX ERGO, which relaxes muscles 20% more

Trackballs are no longer mainstream products as m...