Google is always updating its algorithm. After the app is put on Google Play, what keywords will affect the google play ranking algorithm for your app? Here we discuss the keyword evaluation areas and optimization points involved in the search ranking algorithm in the app store.

Algorithm update about keyword search function and keyword correlation judgment in the search box area.

Previously, Google’s feature updates are aimed at improving relevance of apps returning for so-called "broad" searches, or non-app name searches like "horror games" or "selfie apps." Per Google's words, about 50% of play store searches are broad, and:

"Searching by topic requires not only simply indexing the application by query terms, but also understanding the topics related to the application. Machine learning methods have been applied to similar problems, but success largely depends on the number of training examples To learn about apps. For some popular topics like "social networks", we have many tagged apps for learning, but most topics have only a few examples, and our challenge is from a limited number of The training examples learn and expand to millions of applications covering thousands of topics, forcing us to adapt to our machine learning technology."

Google’s article explains that when they first tried to build machine learning algorithms that could provide good results for these wide-ranging searches, they used deep neural networks, but the results were not as good as the new application discovery they wanted, but they were produced over time. The same application responds to widespread searches, not new applications.

Google’s new attempt is to make this process more like the way humans learn and understand language and word associations. This new attempt utilizes the Skip-gram model, which can predict related words given the input words. Google's new model creates a so-called "classifier" for any given word to create a list of many classifier relations, and finally create {app, topic} associations. In the latest update, Google will also rely on non-machine learning efforts by allowing people to evaluate the quality of the results.

Click "Learn More" to drive your apps & games business with ASO World app promotion service now.

 

Skip parsing

According to the Tensor Flow document, on the left are some example relationships between words, which are determined by Skip-Gram analysis.

Google’s goal is to create an algorithm that can generate a reasonable relationship between keywords (such as {photo} and {share}), and by studying application metadata and user interactions, to generate the best results for a given keyword. Related applications, even if the application returned is brand new. In addition, Google's algorithm must be able to learn new words (for example, selfies, flick, etc.), and be able to establish new associations with these words and other words and applications.

It seems that despite some premature generalization issues, Google is still working to improve the wide range of search results for Play Store users. It is interesting how these changes play a role in the keyword ranking (and download) of all Android apps.

Bottom line: Since 50% of Play Store searches are classified as "broad" (e.g., selfie apps) rather than app names, Google uses machine learning plus manual input to improve the app keyword ranking algorithm when users use broad searches. The ability to return to related applications is used to discover new applications. This may mean that the Play Store keyword ranking is about to change significantly.

Google Play organic search keyword insights

Next, let’s explore more information about organic app marketing, Google Play search keywords and share our relevant insights.

1. Google Play organic keyword data analysis

When optimizing based on data, it is more prudent to first evaluate the data as another key point in the overall plan, and to understand the normal major decisions before making natural search keywords.

First of all, a lot of Google Play Console search data is hidden in "other", the latter is very opaque, and may hide the long tail phrases composed of single words, thus distorting the total contribution margin of a single word; when pressing ARPU/reserve. This is especially dangerous when analyzing rates. Expanding the date range is a way to learn more about "other," but there are still a lot of words hidden in this bucket.

Second, the data is not broken down by country, so it is difficult to clarify regional trends, especially when considering the common language used between countries. This may be an opportunity for ASO tools to provide some type of NLP mapping to suit the needs of the country/region, but this will be an imperfect method and may lead to a decrease in the accuracy of regional analysis.

A safe place to start using Google Play’s organic insight data is to record search term data within a few weeks and redouble your investment in optimizing search keywords, which are always present in the visible word list week after week Strong influence. By evaluating whether your target keywords appear in this list, natural search insights are also a great way to validate your existing ASO strategy (but beware of the "other" category).

2. The keyword grouping of apps and the location of related apps are of great significance in Google Play ASO

Due to the high number of installs from the Google Play browser, the success of ASO is closely related to the apps you find for the correct keyword categories and related apps, and even more important than the displayed results. On the right search keywords.

Unfortunately, although Google provides new visibility for the natural insights of search keywords, Google does not provide a commensurate granularity to explore organic traffic, such as keyword groupings or related apps that cause your app to be viewed/downloaded . In view of the algorithm-driven (ie constantly changing) nature of exploration, it will be a daunting task for ASO to maintain focus on suggestions/related applications and keyword grouping; however, it turns out that these data are correct for ASO Optimizing its Google Play ASO strategy is crucial. At least the success of metadata optimization and UA targeting conquest/app (to increase the likelihood of your app showing as a suggestion/relevant app) can be tracked against the overall trend of store listing visitors and viewers and installers.

 

3. Google Play random keyword insights should be evaluated next to keyword ranking

One of the challenges of reading organic keyword search insights is that during the time period you are analyzing, the conversion rate may and will fluctuate based on your application’s ranking of keywords. If you don’t track keyword rankings alongside your organic search keyword data, the insights you gain may be out of context and jeopardize your decision.

For example, seeing a small number of installers for keyword search may cause ASO to deprioritize the keyword; however, if the keyword ranks 100th and attracts hundreds of downloads, it is actually May be a great keyword to continue to optimize.

4. For large applications, installing contribution margin tilt in Android is more than exploring on iOS

The last most interesting finding is that by randomly sampling apps, we found that the Play Store (organic) exploring the source of large apps usually generates higher install traffic than searches. In some cases, the installs generated by the exploit are 100-300% higher than the installs from the Play Store (organic) search source.

This is very different from the trend of the iOS App Store. In the trend of the iOS App Store (except for the "strange Today" application function), the "App Store browse" source type provides much fewer application units than the App Store search.

There are four main points:

1) Both Apple and Google are interested in controlling the discoverability of applications that have been discovered to attract user interest (ie, high download speeds, high conversion rates, but also ratings/retention rates/revenues).

Neither Apple nor Google seem to care about smaller apps (unless they want to make money from UAC or Search Ads Basic).

2) Google proved that although Apple did its best to update iOS 11 (for example, editorials, "Today" tags, split games and applications, application categories, etc.), Google is better than Apple in applications (especially large applications). (Program) has more control over the discoverability. ). At this point, Google is also more willing to play its role in the pursuit of control. For example, the Google Play Store includes powerful keyword grouping and programmatic suggestions for applications in the application/game view, and nearly endless scrolling, while Apple truncates its application/game functions to support more user-friendly Experience, and the "-y" design style.

3) Perhaps most importantly, as the larger budget releases more browsing/browsing returns, the success of ASO continues to follow the path of "spending money to make money", which accounts for an increasing share of new downloads and searches Big.

4) The last point may involve the whole macroeconomic issue (Eric Seufert?), but one of the reasons Google sees success here may be due to its experiment in redesigning Play Store UX.

For Apple and Google, over time, as the two companies continue to optimize their control of discoverability (and their own checkbook), the percentage of downloads from the browser/browser may increase.

5. The conversion rate of Google Play Organic Explore is much higher than that of iOS App Store browsing

The fourth point is the final discovery. This is the final discovery of the fourth point. This is that the conversion rate of browsing resources from Google Play is not much lower than that of Google Play search. In fact, in some cases, we found that Explore has a higher conversion rate than search. In the case we have seen, the retention rate and ARPU also seem to be strong.

The conclusion drawn from this discovery is that it turns out that Google’s Play Store app discovery algorithm is the same as identifying Google’s original innovation: keyword search, which can identify users who need certain applications, or is close to it.

In view of this, the combined advantages of UAC paid application discovery and Google Play Store browsing discovery may eventually become the turning point for the company to confront another opponent: Facebook. Although forcing app advertisers to use UAC is a huge sum of money for Google in many ways, they worry that Facebook has achieved great success in attracting mobile marketing budgets, but this is a pre-emptive action and it gives Google more time (And data) to train it. When Facebook’s mobile marketing prowess takes off on another “S-curve” with its deployment, value-based similarity and event-optimized campaign positioning in the industry, the algorithm will take off become better.

Google's machine learning algorithm has the unique advantage of learning from organic discovery and paid discovery, which Facebook does not have, and by imposing UAC on advertisers, Google's algorithm learning speed has doubled and the catch-up speed has increased. Even surpassed Facebook. In addition, by training users in the Play Store to click on relevant/suggested apps (i.e. "explore"), Google has expanded the placement of UAC to more locations in the Play Store (i.e., "explore"), thereby increasing Lock in revenue-driven behavior.