LINE Careers

About Work

Q Please introduce LINE's Data Insight Team and your responsibilities.

The Data Insight Team at LINE supports all stages of service and product planning, development, and release by providing insights through data analysis. Depending on the service domain, the metrics and frameworks for analysis vary. Our team, which is part of the organization that plans and develops services related to LINE Messenger, primarily focuses on analyzing user experience and service growth. Our responsibilities are divided into five main categories: 1) Data Product: We standardize and visualize data according to key KPIs and metrics, developing dashboards to enhance data accessibility and support data-driven decision-making. 2) Product Analysis: This includes exploring planning clues, setting KPIs, conducting A/B tests, analyzing post-release performance, and enhancing services. We also perform target user analysis and user clustering. 3) Strategic Analysis: We conduct analyses from a more corporate-wide perspective, supporting strategic decision-making and analyzing the contribution of each service's KPIs. 4) A/B Testing: We develop guidelines and platforms for A/B testing to create an experiment-friendly environment and conduct training sessions to increase awareness and understanding of experiments among team members. 5) Data Management: We design and manage the logic for aggregating metrics that are commonly managed across the company, such as designing the logic for how user time spent on LINE Messenger is calculated. Apart from strategic decision-making analysis, service domain-specific analysts handle these responsibilities cross-functionally. Personally, I analyze overall KPIs for the LINE app, revenue, and specific services like VoIP (voice and video calls), LYP Premium (a monthly subscription service for LINE/Yahoo/PayPay), and OpenChat, which allows real-time conversations based on common interests.

Q What’s your usual work schedule like?

Because of the LINE Hybrid Work system, I work remotely most of the time, which means my commute is very short—I can be at work in just 10 minutes after waking up and getting ready! However, it sometimes takes me a little time to fully wake up. So, I start my day by tidying up, making my bed, airing out the room, and doing some stretching to clear my mind. Then, I make a cup of hot coffee and sit down to start my day. The first thing I do when I start working is check my calendar and go over the tasks for the day. Working in a hybrid environment allows for flexible and efficient meeting scheduling, but it also tends to increase the number of meetings, so I make sure to check them in advance. I handle tasks that require more focus, such as designing queries or analytical frameworks, in the morning. This also includes interpreting analysis results and deriving insights. In the afternoon, I organize communications with other departments and attend regular company-wide meetings for various services or projects. Analysts often share their findings in meetings, ranging from the individual who requested the analysis to project teams and even the entire division. Recently, I've been involved in an A/B test to see if adding an LYP badge to the LINE Messenger 'Font' settings page positively influences subscriptions to LYP Premium. My role includes designing the test (setting metrics to verify hypotheses, calculating sample sizes), conducting data QA and monitoring for the beta test, and supporting decision-making after the experiment. We've had to adjust various test parameters during the beta/RC test phase due to missing logs from the test devices, which has been a significant part of my recent work.

Q When working on international user cases, are there any considerations or preparations you like to focus on?

When analyzing user experiences, it's crucial to understand the users' lifestyles, as data interpretation often involves inferring the intentions behind their actions. Our User Research Team regularly shares reports to help us understand the culture and characteristics of users in different countries, and I make it a point to read these documents. We also conduct many user interviews to better understand different cultures. When reviewing new planning elements, we often combine data analysis and user research to gain insights. While user logs can show objective patterns, understanding the reasons behind these behaviors can be challenging. We often supplement this with methods like focused group interviews. For example, when planning the new “LINE Puri” Photo Booth feature, our team analyzed target users by age and their patterns of using VoIP services to identify which features were popular. User research helped us understand if there was a need for a Photo Booth service among the target group and why certain features were used more frequently. Seasonality also significantly impacts our metrics, so we make an effort to be aware of when schools are on break, holidays, and any significant events in different countries. For instance, OpenChat, which has a high proportion of teenage users, tends to see increased activity around the start of the school year in Japan, Thailand, and Taiwan.

Q What's your focus when collaborating with other roles, departments, or globally?

As a data analyst, collaboration is a significant part of my role. While I initiate some analysis projects independently, many are conducted in collaboration with the planning teams. Our role in these collaborations is to validate hypotheses with data, particularly during the planning or enhancement phases of services. The first key point in a collaboration is to clearly define the hypotheses we want to verify. Clear input is essential for clear output. It's up to the analyst's skills to determine which data and frameworks to use for validation. I focus on understanding the planners' concerns and determining the best metrics to test the hypotheses. The second focus is on clear communication. When collaborating with different departments, there can often be differences in the language or terminology used. It's crucial to address any misunderstandings or ambiguities in real-time to align perceptions and avoid revisiting issues later, which can lead to wasted resources. I always try to keep a history log, especially during the requirement-setting process for analysis metrics. If there's any ambiguity or room for multiple interpretations, it can significantly alter the analysis results. For example, "Chat users" could mean either "visitors" or "users who sent messages." Clarifying these criteria is essential for accurate analysis and mutual understanding.

Q Among the tasks you've undertaken thus far, what do you find the most enjoyable?

Personally, I found the OpenChat User Segmentation analysis to be particularly fascinating because it provides many interesting insights into user patterns. User segmentation involves clustering users based on their usage patterns or characteristics to increase homogeneity within groups and heterogeneity between groups. This allows us to analyze demographic and usage pattern characteristics, offering a chance to see real user behaviors and creatively interpret these observations, leading to fresh insights. What made the OpenChat User Segmentation analysis enjoyable was the clear distinction of characteristics between segments. Typically, user clustering often employs *machine learning unsupervised learning algorithms like k-means, but a common challenge is the difficulty in finding distinct differences between user clusters. To overcome this, we developed a statistical technique for segmentation. We selected four key features (main page visits, chat visits, number of messages sent, and number of days visited OpenChat) and performed segmentation based on combinations of heavy/light users for each feature. This process enabled us to derive insights for strategies to enhance service engagement and loyalty for each segment, ultimately contributing to the formulation of service growth strategies. Additionally, I developed a dashboard to check the volume of each segment and their key metric performance, which could be used in actual marketing and strategy formulation. This project stands out in my memory because of its direct application and impact. *Machine learning unsupervised algorithm (k-means): An unsupervised machine learning algorithm that groups data based on similar characteristics, without specific labels.

Challenges and Failures

Q What does the word “challenge” mean to you?

To me, a challenge means making small efforts to be better than yesterday. This is especially true in the analytics field, which requires extensive expertise across a broad range of areas. Not only do I need technical skills related to data handling, but also a deep understanding of data structures, data platforms, statistical and machine learning methodologies, various business domains, and even data engineering and development areas to produce sharp analytical results. For me as an analyst, expanding into these diverse areas, enhancing my expertise, and growing my capabilities represent my challenges. Moreover, maintaining a mindset of continually improving and not being overly fixated on outcomes is crucial. Striving to excel can naturally bring stress and feelings of guilt when things don't go as planned. Becoming consumed by these feelings can sap the strength needed for new challenges. Therefore, taking opportunities with a light heart and not blaming myself too harshly when results aren't favorable keeps me passionate and driven.

Q Has there been a particular situation or turning point where you feel you grew significantly through your experience at LINE?

After joining LINE, the scope of my analysis expanded from service-level to company-wide issues, requiring me to handle a broader range of business problems and to study and apply suitable analytical methodologies. One example is the LINE Revenue Contribution Analysis I conducted last year. The context was to explore whether the current LINE KPIs and Service KPIs were suitable for monitoring revenue growth or if better metrics were needed. Contribution measurement cases are rare, and quantifying the contributions of various scaled features required a new methodology. I utilized a machine learning model to solve this problem. We modeled a classification model to predict the occurrence of user revenue using each LINE KPI or Service KPI as input features, and developed a logic to quantify contributions by translating feature importance values from the model predictions into contribution metrics. This project, which involved solving a complex problem creatively without pre-existing references, was a significant growth opportunity for me.

Thoughts on competencies

Q What abilities and qualities do you believe are necessary for delivering excellent performance in your role?

I believe it’s crucial to be meticulous to ensure data integrity, and to have logical thinking based on data literacy and creativity. Handling data and BI tools, along with statistical knowledge, are skills that are more explicit and can be quickly learned with targeted effort. Firstly, since analysis involves building logic and persuading based on data, the accuracy of the data is critical. If the data isn’t precise, it can lead to loss of trust. Therefore, analysts spend considerable time ensuring this accuracy. Even a single error in the data can lead to incorrect conclusions, and because numbers are so sensitive, maintaining data integrity requires constant vigilance. Secondly, since data represents facts, it's essential to interpret the underlying meanings behind the numbers, not just list them. This involves seeing the interactions and causations between various data points. For instance, in a *cohort analysis of users leaving a particular service, we found that most were under 24 and had joined through social networking channels offering incentives. High cancellation rates on the day of receiving benefits suggested planned departures, not dissatisfaction. This insight led us to enhance the benefits that were already popular among this age group and expand their experience to other benefits as well. Lastly, the process of storytelling with data is logical but also requires creative thinking. Whether setting criteria during user segmentation or applying analytical methodologies, a bit of creativity is necessary. Even if creating a new methodology isn't possible, applying existing methodologies in new domains or combining them to create new frameworks is an approach that can be taken. *Cohort analysis: A technique that involves analyzing a group of users who share common characteristics or experiences over a specific period, to study patterns of attrition and service usage over time.

Special experiences at LINE

Wrap-up

Q Do you have any goals you personally want to achieve at LINE?

My goal is to make data more accessible to all team members and establish a data-driven decision-making process internally. I aim to ease the discussion, decision-making, and improvement processes based on data. In addition to our current data product and product analysis tasks, I also plan to solidify the A/B testing process. For FY23, we've laid the groundwork with processes/guides, training sessions, and platform development. In FY24, we plan to conduct company-wide basic courses on A/B testing and focus on enhancing these processes and platforms to firmly establish and spread a culture of experimentation.

Q Any message for those interested in this position?

LINE offers an exciting and challenging environment for analysts, where you can analyze traffic from a massive global user base in a free and open analytical and development setting. Not only that, but you'll also have the opportunity to grow alongside wonderful colleagues who provide positive stimulation. We look forward to your continued interest in joining us!