Host: Hello again, we meet again, today we have invited Yang Xun, a CDA licensee to interview us, Yang Xun is currently working as an IT product manager in a leading insurance group in China, welcome Yang Xun, can you briefly introduce yourself.


Yang Xun: Hello, my name is Yang Xun, I graduated from Beijing Jiaotong University in 2020, and my job is mainly responsible for some product design work of the operation module.



Question 1

Let's go straight to the topic of today's interview, can you introduce what kind of products you are responsible for in your daily work?



Yang Xun.

We do a data warehouse product, of course, the same type of products in the industry are also called big data platform, data center and other names. These are actually aliases to emphasize the technical or business characteristics of their products, but in essence, they are data warehouses, a subject-oriented, integrated, relatively stable, and responsive to historical changes in the data collection, used to support management decisions.

On top of this basic concept, the use of some distributed technology to support the processing of a certain amount of data is big data, the construction of some common upper-level data applications is the platform, to maintain the stability of the underlying system data under the premise of supporting the flexible and changing business needs of the front office is the middle office, but the most essential things are still the same.

Because we are an internal product, we don't have a fancy name, but mainly to solve some of our internal practical problems. It collects data from various business systems together, and can perform some complex cross-business and cross-system data statistics work. At the same time, as a data hub, it provides a standard and secure data interface, which effectively improves the efficiency of data interfacing among various business systems. After breaking the data chimney between systems and allowing data to flow fully, the value of data will be further brought into play.


 

Question 2

As a product manager, you have data analysis skills, do you think it is "empowering" or necessary?



Yang Xun.

There are many types of product managers, and there are many types of business involved, so the level of data analysis skills also varies greatly, so I'm afraid it's hard to accurately describe this question by simply choosing one or the other.

I think from the perspective of data analysis skills, product managers can be roughly divided into three categories: general B-side products, C-side products, and data products.

The general B-side products are facing enterprise customers or their own company. However, the user volume is limited, so it is generally difficult to conduct effective user behavior analysis. A more efficient and direct approach is to communicate with some core user representatives to understand their business processes to form product solutions and collect their opinions to improve and optimize the product. This kind of product manager has low requirements for data analysis skills, and can master some basic skills of Excel is enough.

The second category is C-side products, C-side products tend to have a wide range of users, if not frequent user data analysis, it is likely that after a while even their own users do not know who is. Therefore, for C-end products, collecting user behavior data and conducting multi-dimensional data analysis are important means to understand their user base and generate user portraits. A large part of product requirements and improvement ideas are also based on this. So these products need to master some of the more complex data analysis methods, such as trend analysis, forecasting, clustering, etc., the tools can be used SPSS, Python, R, etc..

The third category is the data products. Because the business itself is to deal with data, so naturally the data processing, analysis and understanding of the ability to require relatively higher. Compared with other product managers, data products need to have basic data analysis skills, but also the use of mainstream SQL or NoSQL database mastery, the common big data platform architecture and its principles have an understanding. If you compare data to fish and data analysis ability to fishing ability, what data product managers are doing may be more like making fishing nets or fishing rods.

As to whether it is empowering or necessary, if we take the CDA's ability level as the reference standard, I think for ordinary B-side products, it is a good idea.


I think that for ordinary B-side products, about half of the content in Level 1 is essential and the rest is enabling.

For C-tier products, Level 1 is all essential, and the rest is empowering.

For data products, all the content in Level 1 and some content in Level 2 are essential, and the rest are enabling.

Of course, I roughly divide the product manager into three categories, there must be a lot of places that are not rigorous, and do not exclude some special circumstances have higher requirements for data analysis capabilities. The above is only a general summary based on my understanding, I would like to offer you criticism and reference.


 

Moderator: Is it possible for me to understand that CDA Level 1 is a necessary skill for a data analyst?

Yang Xun: Basically, I can understand it that way.

 

Question 3

What are the data analysis methods used in the usual work as a product position? What tools do you need to use?



Yang Xun.

Because it is not a professional data analysis position, so some data-related work is usually done in the workplace mainly to serve the product design.

In the process of data warehouse construction, we need to use various databases frequently to understand the structure and shape of data sources, and we need to use dimensional modeling techniques to classify and manage data on various topics.

In the process of building the index system of operation module, it is necessary to use data presentation technology, be familiar with the characteristics and applicable scenarios of various types of charts, design logical combinations of indicators, choose the appropriate form of data presentation, and finally present the results on the data dashboard to support management decisions.

The most commonly used tool is Excel, but occasionally Python or spss is used for simple data processing or more complex statistical chart presentation.

 

Question 4

The feeling I had before about the insurance industry was that we were handing out flyers on the street, and people were avoiding them, and we basically don't see such flyers now. What is the reason for such a change?



Yang Xun.

In fact, the domestic insurance industry gradually became active from the 1990s, and from then on, it gradually came into the public's view. Many insurance companies, including our company, were established during that time. In the beginning, there was no good experience and method, which was to sell insurance as ordinary commodities without any difference, so it was difficult to accurately match the real needs of users. In addition, most people lack understanding of insurance, just let me pay money, not let me feel what I actually get, so it is easy to produce the illusion that insurance is a scam. I think this is the reason for the inherent impression of the insurance industry that you just mentioned.

Next, I would like to talk about the changes. With the development of the past 20 to 30 years, more and more people understand insurance better, and insurance is actually different from ordinary consumer goods. Unlike ordinary commodities where buying or not buying depends on whether you like it or not, buying or not buying insurance depends more on one's risk appetite and actual needs. Therefore, more insurance brokers will emerge to choose the most suitable insurance portfolio for their clients. Compared to the traditional agent-based sales model, it is clear that the broker model is more responsive to the needs of the user. This is one aspect of the business that has changed, as reflected in the changing needs of customers.


The second aspect is that the industry itself has been thinking and exploring for a better business model for such a long time. From the beginning of simple imitation and learning from Western models, to the recent years of active innovation, the shape of the industry and the sales model itself is constantly changing. Like our company's first efforts to create the three closed loops of "longevity, health and wealth", we have shifted from the traditional pure insurance business to the business model of building a large health ecosystem. So far, our high-end retirement communities and hospitals have been laid out in 26 cities across China, which is the leading position in the industry. The business model of the retirement community simply means that you can sign a policy called "Happiness has a contract" and get the right to stay in the retirement community. In the case of Yan Yuan in Beijing, the cheapest policy is $2 million. So from this aspect, sometimes we feel that we see less and less insurance ads, not because insurance companies don't advertise anymore, but they are more targeted. Like me, I still have to work hard to receive the advertisements of Happiness.

The third factor, which I think has the most obvious effect, is the wave of mobile Internet that started around 2010. Under the trend of mobile Internet, not only the insurance industry but also various industries are rapidly changing their service models, and major traditional enterprises are seeking digital transformation to adapt to this trend of the times. People are increasingly inclined to centralize various services on their cell phones. Taxi, shopping, ordering take-out, mobile payment and the popularity of digital government platforms in the past two years have moved many things that were offline to online. Our company also set up a subsidiary of Taikang Online in FY15, which is specifically responsible for Internet insurance business. So from this aspect it is because the position has changed, moving from the street to the internet.


In a nutshell, there are three reasons for this. First, the demand has changed as people's understanding of insurance has deepened; second, the industry's own exploration of business models has led to a change in supply; third, the connection between demand and supply has changed due to the development of the Internet. The above three reasons together have led to the current result.

 

Question 5

The analysis is very thorough, just mentioned the development of the Internet has led to changes in the link between demand and supply, how do you think we can do a good job of data burial?



Yang Xun.

Speaking of buried points, we just made a simple buried point tool last year. In fact, the company has purchased commercial buried point systems like TalkingData and Shen Ce. The feature of these systems is that they are very powerful, but one of the problems is that the cost of buried points is much higher. Because your system wants to carry out more dimensional and complex analysis, it means you need to provide more information when you bury the points, and more specifications and constraints. This is also some kind of conservation law, the more you want the more you have to pay. I've been using Shenzhe for a while, and the buried work itself does take up a lot of time in the iteration, sometimes even need to arrange the whole iteration to buried for the pages that didn't do buried or didn't buried well before. Then many of our internal ToB systems actually do not have complex buried data analysis needs, and are more concerned about the most basic data indicators such as the utilization rate of functions and page access. So we built a buried tool last year based on logging platform, ETL, data dashboard, and configured our encapsulated listening function into the front-end service, and then we only need to name the page tags according to our set of specifications to achieve buried points, which basically does not generate additional development volume. Later, many internal systems that were not buried also began to be buried.

Because of the buried point so above a brief description of the buried tool, not necessarily the more complex the better, but also depends on the specific needs of the scene, the complexity of the function and the buried point convenience is not both. And on how to do a good job for their own system data buried this issue, I think we still have to start from the product goals.


In the product goals guided by the formation of buried goals. For example, an e-commerce product, your goal of a certain function is to make users more click on the product details page, more orders to pay, then your buried point is not to be able to reflect how many users clicked on the details, how many users paid, it is best to buried data analysis out of the follow-up how to improve the product can improve the payment rate. This is the goal of your buried point.


Then start the buried point design under the guidance of the buried point goal. The core principle is to know that "all indicators are one-sided", around a business goal may require multiple data indicators to form a combination of indicators, in order to more complete response to the actual situation. A single metric is often not convincing enough.


Finally, after generating the buried data, effective data analysis is conducted as needed, and problems with the buried design are identified and continuously improved in the process of using the data. If these three steps are done, I think this part of the data buried work is basically done.


 

Question 6

Last question, as a data analysis product manager, do you have the most important thing you want to spit out?



Yang Xun.

Frankly speaking, I can't think of anything to complain about. I think when you encounter a problem, it is a priority to look into yourself to see what you can change, and to face all kinds of problems positively and optimistically, which may be a more effective way to solve problems. Trolling will only give a feeling of powerlessness.

And the privilege to be invited to participate in our exchange, a lot of big words, inside already feel very ashamed, more can not think of what to spit. I feel that when you are young or input-oriented, listen more, read more, learn more, do more, less trolling ~

 

Conclusion.

Today's interview is by far the longest interview, chatting very happy, Yang Xun as IT product manager combined with their own work content to share their experience, in the company also experienced many iterations of the product, business model changes, and even links with customers have changed, but also opened a new subsidiary dedicated to docking, here is undoubtedly not the Internet has played a huge role in promoting, once again Thank you Yang Xun to participate in the CDA licensee interview, we will see you next time!