Zigzag has data without DB - ZDNET KORE

Kakao style zigzag is the most successful female shopping mall app between the years. With a 5,000 different women's fashion shopping malls, the mobile app that recommends styles per person, with a mobile app that recommends styles, the zigzag trading liquid exceeded 750 billion won, and this year is towards 1 trillion won. Up to June, the cumulative app downloads up 30 million, the moon, 90 million won, and the monthly user number (MAU).

Zigzag has led to success in data-based business. We have developed a personalization recommendation algorithm that analyzes the use patterns by analyzing the user's preferred Sooho Mall, Brand, Interested, and Purchase History. This algorithm provides a customized recommendation for users perfect for personal tastes. In December 2017, a business model introduced a personalized advertising system. We analyze behavioral patterns with user big data, which accumulated from the earliest of the service, and recommend the goods depending on individual visits and purchasing history. With personalized recommendations, users are recommended for their taste without a sense of rejection, and the seller can conduct advertisements for target customer groups to create maximum effects and minimize advertising resources.

The organization that supports the data-based business of this cacao style is 'data group'. Data engineering teams, data analysis teams, and data science teams are consistent. The Data Science team develops data models such as personalization and recommendation, and the data analysis team uses internal data and log data to analyze app usability, business, and so on. The data engineering team will build a data platform, which is the basis of all activities in the team, such as data science, data analysis, and is responsible for engineering to manage large data storage, processing, utilization, and so on.

The Cacao style data group is the so-called 'function organization'. The organization of Kakao style is largely divided into 'function organizations' and 'destination organization', and the objective organization is made up of project form, and is a team of planners, developers, and more. On the other hand, the function organization supports a team of warriors by collecting an expert related expert for a particular feature. The data group consists of data-related experts.

"If you create a recommended model in a data group, you will be able to apply a recommendation algorithm through engineering to each service", "If you create a recommended model in a data group, you will not be able to apply a recommendation algorithm to each service." In the event of a respective? I explained the inside of the inside and use it in a transcriptional form. "

He said, "It was scattered in each of the 100 units, when it was a company of 100, but it has been scattered in each objective organization, but instead of increasing the data this year, he gathered to one team as now, and to emphasize the efficiency," The organizational composition has been dependent on the size of the company and added that it will vary flexibly according to the organizational situation in the future. "

Currently, the Cacao style data group was done with 14. There are four people on the data Science team, 7, and three engineering teams.

Cacao styles actively utilize the services of Amazon Web Services (AWS) to Data Platform Construction. It is unique that it did not separate the analysis DB separately from the service operational database (DB).

"By default, all data is stored in Amazon S3, and all AWS services can be written from the moment of real-time data or operated DB Damage S3," Modeling is a Sage Maker, Dashboard Quick Site, Data processing processing is focused on using EMR, etc., which is focused on free and selecting the service to deal with data. "

"If the Data Engineering Team Reader is a bigger SQL DB," said Data Engineering Team Leader, "I am working on a DB of SQL DB" and "I have been building a data lake to save all data to AWS's S3 storage." .

Maintenance The leader said, "The actual retention data amount is a 250 terabyte (TB), and only a few billions of days, which will be logged in S3, and the AWS glue service is interfaced to write in the form of S3" "This is AWS Athenan Query Services, which has more than half of the Cacao style entire number of people."

The dashboard visually to analyze data analysis is using the 'Amazon Quick Site'. If the Data Analysis Team or Data Science Team creates an Amazon Quick Site view, the company's business team uses this dashboard to decisions.

Recommended elements are using AWS Personal Ryzan Service. AWS Sage Maker is used to create and use a customized synonym. When a large-scale data operation is required, a Big data operation is performed using the Amazon EMR, which is a managed Hadoop framework service.

"Amazon EMR," Amazon EMR could be written when they were in the same time, "said EMR," he said, "EMR is still back, and at the same time."

He emphasized, "I had to expand infrastructure in the past," he said, "In the case of the current platform, it can be done in the form of 10 to 100 times," he said.

Data stored in the data lake does not contain personal information. ISMS-P certified. Personal information is outside the data lake, and the data of the data lake is stored in a state where the individual can not be identified. Without user identification information, recommendation modeling or analysis is working well in the actual service, and it has a lot of interest in the industry in the industry.

The Sales Leader said, "It is not common to receive ISMS-P certification in this way," and "Normal to lock all data to DB and not external. It is an expression that receives and writes it in the service, and we have been saved in a utilized form, but it was as much as possible, but it was not a problem with information security. "

The real-time data stream utilizes Amazon Kinesis Fire Hose. If the amount of real-time data increases, Kineysis Fire Hose is easily expanded and written. The Su Saesung Leader said, "The behavior log, which is loaded billions of figures in a day, is loaded by using Kineysis."

The data analysis team analyzes the performance of the business service and performs continuous monitoring tasks. Amazon Athena has been automated with Amazon EMR for analysis of data analysis, and continuous monitoring is required for analysis that requires continuous monitoring.

Data Analysis Team Park Inseong Reader said, "If you write a query or process that is made with Amazon EMR, when the analyst wants the analyst, the task is daily, and the source data is stacked and the source data is stacked and the logic based business index is created, "If you say that it is called" business index information that is, it is necessary to convey tools, it also requires a dashboard to form a flow of collecting, processing, and delivering with the Amazon Quick Site. "

The Data Science team utilizes services such as posters and recoggings that provide applications available at AWS. Recently, I am using Amazon Sage Maker to create a custom machine.

The Sales Leader said, "If you have focused on creating a performance confirmation and business by applying the AWS personalization recommendation model for a quick action, you have focused on creating a performance confirmation and business, you will now increase and diversify your personalization and recommendations in your organization. To make a custom model to make a custom model, we are introducing a sage maker to increase the utilization. "

"For example, the advertising service is important and efficiency and efficiency. The ROAS indicator predictive model to determine the advertising efficiency of Zigzag is distributed by creating a sage maker. "It is possible to concentrate on the other after 14 days after the actual advertising service provided to the partner, "I have explained.

He added, "Various recommended models, and a variety of machines will be further expanded."

The sector of the most dramatic effects of the AWS has adopted various data related services.

Park Inseong Reader said, "If 7 of the analytical teams did not have a cloud, it would have not been able to support the level of data now" "Analysis is not only data, but also the biggest performance that allows you to connect directly from the service or business stage, "I said.

"There are a professional group to analyze the user's response by providing the user as an AB test, analyzing the user's response daily, and I have a specialist group that finds the user, and to find the missed part or improvement, and it is closely collaborated with it." Commerce In terms of MD and plan implementation, the planning exhibition is to share experience after termination, but it is worried about Daily data and analysis, and it is worried about the MD to improve the product or the concept even if it is a short period of time. "

In the cacao style, data is used for all duties. Utilizes MD, developer, and data, and the PR stalls also cover the data directly to the data group using the platform that the data group is well wiped. The output of the dashboard or the data analysis team can be easily seen anywhere in the office.

The Sales Leader said, "The data is not closed, but someone can be aggregated if there is a deal in the company." This environment is natural in the cacao style, "I have not emphasized."

The data in the data group is provided as possible to partner sine shopping malls beyond the utilization of the company. Even if you do not look directly in your internal inside, you will see data to help you operate your partner or help you grow your own shopping mall. Or regularly, data reports are published and distributed to each shopping mall.

"The report is not to make the operations of multiple singers, a performance report, but a regularly arranged indicators in the usual prepared infrastructure." The side is as follows to support as much as possible. "

"The representative case is" 1 "as a" 1 recommendation ", and when a customer is hesitant or hesitates, it gives a nuts, or in contrast, when a jigzag point shopping mall partner issues a coupon to a particular customer, it is efficient "I am helping to go well with it."

Cacao style data groups have asked for the reason for chosen AWS among several cloud services.

In this paper, "Zigzag has been very complicated," Zigzag has been very complicated, but the other cloud also wrote a little, but the AWS service has made a satisfactory service without having a great difficulty, "said the AWS service is a technical limit that is difficult to unwind in the current paradigm. When I always have the latest technology to introduce the latest technology, it will open the ticket to open the Ticket, and communicate with the actual AWS's service developer, and answer the technical trends or our questions in AWS. "

"AWS is a" AWS to designate the enterprise support target customer, and the person is resident on the sleeves of the CacaOS style data group, "said the person in charge of the Cacao style data group," and "AWS" and "AWS are resident on the sleeves of the Cacao style data group" and " He added it to me. "

Cacao style data group is not completely excluded DB. It has a policy that reviews the most appropriate technique to solve difficulties that occur on workloads. First, it is said that DB technology is set and does not fit the workload there.

Maintenance The leader said, "If you need a quick response, you need to write Elastic Cassie, Redis, and you need to write Amazon Aurora," There is a result of what the data lake has been calculated, and you do not have to write it in the analysis, When passing to write to write, it also introduces Amazon Aurora, depending on the data type and circumstances, and writing Redis. "

There are a variety of opinions in the questions where you have a service that you want to request to AWS. Yu Ji-hoon said he would like to be able to use the Apache Air Flow Right Amazon Managed Workflow (MWAA) in Seoul Region, which provides 'Air Flow', a scheduler developed by Airbnab. He also made a comment that I would like to support the graph in 'AWS glue'. The sedimentary leader wanted data discovery to be easy to find the location of data from the data lake. This idea was conveyed immediately to the host office through AWS Korea staff, which was expressed in the interview.

Cacao style data groups are further enlarged for personalization areas. Personalization is not only product composition, as well as user-specific UI and customized content recommendations. The Sales Leader said, "We will provide personalized UI and content, and will also be made of data projects that apply the insights and successes obtained from zigzag data to another service of CacaOS style."

"The AWS Red Shift is a data warehouse quickly with a data warehouse, and it is trying to use this DW to provide a variety of insights to a variety of insights," and said, "The existing method is limited by the workload I put the red shift, "I will invest for a partner because there is no limit."

The Sales Leader said, "Increase the absolute amount of indicators to provide partners. Emphasized a customized consulting on a 5,000 seller individually, "he said," he said, "It is said that it is not a" consulting firm, ", but it will not be able to attach a consulting, but it will proceed in a direction to provide a consulting or data to provide personalized consulting or data."

Park In-Sung Leader said, "We plan to provide data to the seller site according to the scope and permissions of the dashboard," says, "" I think that "the troubles of the customer and the troubles of the customer should go together."

Data-based business is a direction to be adopted by many companies. Startup, large enterprises, etc. are interested in using data, regardless of size. CacaOS style data experts have made an advice such as setting objective goals, to invest in enterprise-wide data accessibility, to build a data collection system that can be flexible.

"To work on data based on data basis," he said, "If the target is difficult to objectify, it is amazing and the data is far from the decision," he said.

"To ensure that quantitative goals are well maintained and linked, enterprise-wide data accessibility, and systematically, they should be able to view the rights, and the data can be viewed anytime, and the data should be seen." The time to be required is always emphasized. "

"One day, someone has a dashboard, when someone has a dashboard, sending inquiries to the analysis team, and it is important to make such a flow," he said, "he said," To this end, we have to organize a well-delivered flow from Data Lake " Added.

Kang Wong Suk Engineer said, "It is better to start and start flexibly from the point of view," he said, "I want to eat more, and I want to eat more, and I want to do everything data."

He said, "I have not been able to put it in RDB because I have not shown a few years ago, because I have not been developed in RDB, now I have a lot of excellent data services in the cloud vendor," he said. " I added it, "he added.