Product | technology drops
The author | Song Shijun
The author of this article is Song Shijun, head of data science Department of Didi Chuxing. He once worked in core departments of Facebook and Google, and is a well-known Data analysis director of Chinese. Mr. Jingshijun is authorized to share with you here, hoping to let you understand the core value of data analysis for a product, whether DS, R & D or product students grasp the data. I hope I can help you.
— — –
DS is a new function in the market that has emerged in recent years, and we are still very young compared to r&d, algorithms, products, operations, etc., which have been evolving for 20 or 30 years.
On the one hand, the development and demand of this function can be seen from the supply and demand of talents in the market, but on the other hand, like any new thing, this new function also has many challenges. Today, I want to talk about how I view the function of DS and our development direction.
First of all, we should make it clear that DS is not a “necessary” function of a company, but there will be the inevitability and rationality of THE emergence of DS in the development and expansion of a company. We are like the aim on a gun. We can shoot without the aim, but the aim makes the gun more useful. The company does not have anyone to do data analysis, and it can still run in the short term, but in many places it will not run very well; If one day the company do data analysis people disappear, the company will not collapse in a short time, but a longer time will certainly have an impact.
When we are not “necessary” functions, we have to ask ourselves “who is DS?” “What does DS do?” “What is the value of DS?” “Where is DS going?”
▎Who is the DS
In psychological terms, this is the DS id. We are a group of people who have received professional training in the field of quantification, and hope to use our quantification ability to mine useful information in the data, and provide help for business development through these information, while maintaining the neutrality of data.
A function (or position in a company) is determined by what it should be doing, not by what it is doing. So, we describe DS more in terms of what we think we should be doing than what we are doing. For example, many students have such questions as “DS does a lot of fetch”, and even many business partners “expect us to meet a lot of fetch needs”. None of this has anything to do with who the DS is, it just means that we haven’t done our job yet and have a lot of work to do (more on that later).
From an individual point of view, this also means that we look at DS not in terms of the person’s academic expertise, but in terms of the person’s motivation and willingness. There are a variety of data-related functions in a company, and some use data as a means to get business results. Be accountable for business results. These are data operations. Some of them are data as an object of development, responsible for these products that are related to the data, and these are engineering development. Some of them are real-time, online implementations based on data, and that’s what algorithm engineers do.
These are our partners, but we have our own positioning, which is different from these. We should be responsible for the neutrality and science of our work. We need to have a business idea, but we don’t want to be the business itself, we want to be the catalyst for the business.
▎DS do
I summarize what we do, which can be abstracted into three categories: (1) describing the situation, (2) looking for patterns, and (3) promoting improvement. These three things are layer by layer, but all are important.
DS first describes the status quo, which we often call “counting.”
When we can’t even describe the objective situation clearly, we can’t talk about finding rules and pushing for improvement. A lot of work in our work, we do indicators, do data report kanban and so on are in this category. But why do many students have great doubts or feel no sense of achievement about the work of “taking numbers”? I think it’s because we’re taking numbers passively, or we’re not connecting the number taking itself to the main line of our business, we’re just filling in the blanks.
In addition, I mentioned in ten points of data analysis that “what problem to analyze is often more important than what method to use”, and “what number to pick” and “why to pick” are often more important than “how to pick” and “how much”. Most of the time, thinking about “why to take” from a business perspective gives us a stronger sense of value. If we can actively think about “why to take”, we can feel more engaged. It’s a first step, but the value is enormous, and if it doesn’t help the company describe what’s going on, the company is going blind. This first step requires each of our students to have the ability of independent thinking, especially critical thinking.
DS also looks for patterns.
The essence of data analysis is to look for patterns, to look for patterns that are hidden in the data that others have not yet discovered. We talk about statistical extrapolation, causality, growth drivers, predictive modeling, experimental evaluation, and so on, all looking for patterns. These rules are what we call “insights.”
Of course, the rule of gold content is not easy to find, which is the value of our DS. If we can see patterns that everyone can see, then we’re not providing value; Those who can dig deeper and see more essential rules will provide greater value. Therefore, our academic training, scientific methods, practical experience, data sensitivity and so on are all helping us to find value that others cannot see. So I encourage you to describe your work not in terms of what methods I use, but in terms of what patterns (insights) I see. This requires each of our students to have a strong curiosity and firm faith.
We describe the current situation and look for rules, the ultimate purpose is to promote improvement, which is often referred to as impact. I have concluded that the impact of DS can be divided into four categories: (1) improving key metrics, (2) influencing product decisions, (3) influencing operational processes, and (4) creating sustainable solutions.
If we do something that doesn’t directly or indirectly fulfill any of these four categories, we need to reflect on whether we are spending our time in the right way. And what can we do in the future to maximize our output per unit of time invested? Ideally, before doing something, think about (if it’s a passive demand, ask the demand side) how what we’re about to do will have an impact. To achieve these impacts, we also require each of our students to have empathy and business (product/operation/marketing, etc.) thinking, as well as the ability to refine, excellent communication skills, the ability to persuade. Understanding the four dimensions of our influence explains the value of DS’s existence. From a psychological point of view, this is the equivalent of DS’s superego.
▎Which direction is DS going to go
This is rather a question of DS ‘ego.
I have summed up this issue in two aspects: capacity building and culture building. In the direction of capacity building, we need to be strong. We need to be able to do more in-depth analysis, use more scientific tools, so that we can do what others can’t do, and we can see what others can’t see. One thing to emphasize here is that competence is not just technical competence, but also business thinking competence. We organized the Delta program to help students develop this ability. We also encourage students to travel to study groups, rotate jobs, and communicate with senior experts in the group. Improve your ability. At the same time, we encourage you to think about what data can do from a business perspective. Learn more from the thinking mode and perspective of business leaders, and then combine our data accumulation to form our own things.
Just as important or even more important than capacity building is culture building. To change how the environment (colleagues, company, industry) views DS, we need to be firm about how we see ourselves. There is a confidence issue. Our worth is determined by what we do (the self), and this is not dependent on external recognition and approval of us; We want to increase our value, which is essentially how to make what we do more valuable. With confidence, we have the direction to guide how our colleagues and colleagues see us, how we can make sense of what we can do, and how others see us is essentially a reflection of how we see ourselves. If we feel we should count, we count in the eyes of others. If we tell people that our time could be better spent elsewhere in the business, then both we and the other person have the will to do so. And if we can fulfill these through efforts, the other party will confirm our position more and form positive feedback.
The same problems we encountered in the DS team, I encountered at Google and Facebook several years ago, but through the efforts of the whole team, we gradually proved ourselves, established the DS brand and recognition in the market, and were recognized as the benchmark of this function in the market. DS as a function has also gained similar status as engineering and product, and has recently been voted the most promising job on several occasions. This process is gradual, it will take time, and it will require us to work together.
In fact, we are doing the same thing at Didi. DS and data-driven concepts are still in the early stage of development in China, and many things are still at the theoretical and emotional level, which is similar to the state of Silicon Valley several years ago. This is also why students in our department are facing so much confusion, and we leaders need to help them to firm up their direction, because we are the people leading this function in the market, exploring and expanding the boundary of this function, and this process is bound to be challenging. Different from other functions, our leaders and grassroots students are creating the history of this function while doing specific things.
▎Article 10 Data Analysis
Finally, I would like to refer back to my “ten Data analytics” list, which reflects many of the aspects mentioned above:
1. The core ability of an analyst is to think [what DS does]
2. Be responsible for telling the truth and remain neutral [who is DS]
3. Sufficient arguments, rigorous arguments, concise views [to promote improvement]
4. Data before opinion, not opinion before data [what DS does]
5. Don’t make things complicated and don’t be afraid of complexity [who DS is and what DS does]
6. What problem is analyzed is often more important than what method is used.
7. Good analysts give input, not just output.
8. Analysis is of little value unless insight changes something else [non-essential function]
9. If possible, collect data based on questions, not just ask questions based on data [not mentioned in this article]
10. Not all questions can be analyzed. Be open to other points of view [not mentioned in this article]
▍END
Senior data analytics industry leader, former Director of Data analytics at Facebook, Head of Data analytics at Instagram, and Head of Long Tail Advertising growth at Google, he has extensive experience in data applications for Google and Facebook clients, advertising, and social network content production and consumption.
At the same time, didi data Science department is looking for positions of data scientist, data analyst, data development architect, etc. Students with strong interest and achievements in data science are welcome to apply. Resume should be sent to [email protected]. For specific job information, please refer to talent.didiglobal.com/.
