Analytical solutions for smartphone games
G2Metrics custom-made analytical solutions for smartphone games were built and released in March, 2015 on Google Cloud Platform (aka GCP) at StarGarage. We asked Mr. Takuya Sugiyama, CMO, to talk about the content and development of their services.
G2Metrics is an analytical solution for smartphone games. It can generate a variety of reports on things like basic KPIs, cohort analysis, custom reports, game indicators, user retention, real-time analysis for analyzing player activity, managing game balance, improving campaign/promotion efficacy, increasing game revenues, and so on.
By analyzing user attrition or subscriber trends in real-time, we can build a “winning pattern” of subscriptions to the application and support maximization of revenue. Released officially in March, 2015, it has already been introduced and implemented for game titles such as “Defence Witches”, “Pop’n Cube” and the “‘Binoba’ educational game”. Game app developers can use the service for a broad range of activities, and it is free of charge for up to 20,000 MAU (Monthly Active Users). Therefore, any companies who’ve recently released or who currently run any games are welcome to try it out.
Take, for example, the App Store where over 1,400,000 applications have already been released. 430,000 of these come under the gaming category or sub-categories, but 80% of apps are almost never used. Also, only 13% of all apps are able to raise any revenue.
With G2Metrics, we wanted to create a platform that would make games developers’ hard work bear fruit. If you look at the mobile app solutions genre, there are already several major services such as Flurry or UPSIGHT, which are generic services for mobile apps, so although on the one hand they can do a lot of things, when we introduced them, there was a problem with exploiting the collected data effectively.
On that point, G2Metrics is focused specifically on real-time analysis of games which makes it simple to perform real-time analytics of basic KPIs, such as Daily Active Users (DAE), turnover/installation figures, etc. as well as cohort analytics such as subscribed/unsubscribed users, users with a fixed number of friends, users who own specific items, etc. using all kinds of game indicators such as custom reports, user trends during boosting and short campaign periods.
We use GCP services Google Compute Engine (aka, GCE), BigQuery and Cloud DNS. We also use Load Balancer and LocalSSD for GCE. We initially considered using Cloud Storage too but finally we decided to leave the data as it was in BigQuery because there isn’t a big difference in storage charges.
Though we felt that BigQuery has great potential, G2Metrics infrastructure was a big reason for choosing GCP. Real-time analytics is a feature of G2Metrics, so we initially envisaged it as an appliances analysis platform, though naturally the option of using the cloud came up. When we conducted surveys on AWS’s Redshift, BigQuery came up as an alternative and our level of expectation that we could use it to achieve our goals rose when comparing functions and costs. Finally, we participated in the GCP seminar, tried out GCP at Google’s hands-on workshop and once we were confident that it was viable, we decided to develop G2Metrics entirely on GCP.
One good thing is how quickly it activates instances. In the initial release, we’ve been using 13 GCE VM instances but we plan to operate with the optimum number of instances while keeping an eye on the load status. Being able to use a load balancer with no pre-warming is another thing for which we rate it highly. We want to quickly build a service where the load balancer is practically screaming out with the strain from the number of subscribers!
Before we started using it, we felt that the user interface was weak compared to AWS, but we found that it evolves on a daily basis and people who originally invoked the command line can use it with virtually no resistance.
On the negative side, BigQuery sometimes returns unknown errors when we run queries. Although we hope that Google will address this issue in the near future, we can turn a blind eye to it for now considering the current benefits of using BigQuery. Due to the emphasis on real-time capabilities, we are unfortunately constrained by the upper limit on requests for frequent queries run on BigQuery, so we would like this to be addressed flexibly. We anticipate the same for new services such as Dataflow.
Also, we would like to be given road maps prior to Google’s daily release of new functions, which along with the interface issue I mentioned earlier, is a problem unique to Google.
We took advantage of Partner Billing and Single Incident Technical Support as additional services. Partner Billing reduces time spent on accounting procedures and allows payment in yen, so there were no administrative problems related to our adoption of GCP.
In addition, the quick response technical support was a help when dealing with some small technical issues within the short time period up to the release date. The service helped our staff resolve technical questions so I was glad we used the service as it saved me a lot of time dealing with their problems.
What plans do you have to develop your services?
We are aiming to exploit BigQuery’s real-time analytics capabilities to support the profitability of game development companies and become Japan’s biggest gaming analysis platform. We would encourage all companies looking to release smartphone gaming apps to make the most of G2Metrics’ effective game developments and operation!
This is a translation of an article published by Cloud Ace, Inc.
Available online: http://www.cloud-ace.jp/case/detail13/