The most important thing in data analysis in product operation is to have the thinking of data analysis, master the methods of data analysis and use the tools of data analysis. Next, I will explain them from these three aspects! There are many dry goods. I suggest you have a look first. As a product operation, we must consider the thinking of data analysis: where is the essential value of data? What can we learn from these data? What can you guide us to do? Faced with massive data, many product operators don't know how to prepare, how to proceed and how to draw conclusions. The following is 1 the classic five-step idea of doing data analysis: the first step is to explore the business meaning, understand the background and premise of data analysis, and what is the result of business scenarios. The second step is to make an analysis plan, how to split the scene and how to infer. The third step is to split the required data from the analysis plan and really analyze yourself. The fourth step is to extract business insight from the data results. The fifth step is to gain insight according to the data results and finally produce business decisions. For example, an Internet financial website in China, the marketing department has continuous advertising on Baidu and hao 123 to attract web page traffic. Recently, internal colleagues suggested trying to launch Shenma mobile search channel to obtain traffic; In addition, it is necessary to evaluate whether to join Kingsoft Network Alliance for in-depth advertising. In this multi-channel delivery scenario, how to make in-depth decisions? Let's disassemble this problem according to the five basic steps of the above business data analysis process. Step 1: Dig the commercial significance. First of all, we should understand what the marketing department wants to optimize, and measure it with Polaris index. For channel effect evaluation, business transformation is very important: for P2P websites, whether to launch "investment and wealth management" is far more important than "the number of visitors". Therefore, whether it is Shenma mobile search or Jinshan channel, the focus is on how to measure the transformation effect through data means; The operation strategies of different channels can also be further optimized according to the transformation effect. The second step is to make an analysis plan. Take "investment and financial management" as the core transformation point, allocate a certain budget for traffic test, and observe and compare the number of registered people and the final transformation effect. Write down the number of times these people repeatedly buy wealth management products to further judge the quality of the channels. Step 3, split the query data. Because we need to compare the channel traffic in the analysis scheme, we need each channel to track the traffic, landing page residence time, landing page pop-up rate, website visit depth, orders and other data for in-depth analysis and landing. The fourth step is to refine business insights. According to the data results, compare the effects of Shenma Mobile Search and Jinshan Net Alliance, and infer the business significance according to the observation results of two core KPIs: traffic and conversion. If Shenma mobile search is not effective, you can think about whether the product is suitable for mobile customers; Or carefully observe whether the performance of the landing page can be optimized and so on. And need to find business insight. The fifth step is to make business decisions. According to data insight, guide channel decision. For example, stop the launch of Shenma Channel and continue to follow up Jinshan Network for evaluation; Or optimize the mobile landing page, change the user operation strategy and so on. Every time you do data analysis, you can refer to these five steps of product operation. Second, the method of data analysis There are eight common data analysis methods. Mastering these eight methods is basically enough for people who operate products in the middle and low grades. Taking an e-commerce website as an example, we use the data analysis product GrowingIO to quickly collect data, display it clearly and intuitively, and then share these eight commonly used data analysis methods with you. 1. Numbers and trends Looking at numbers and trends is the most basic way to display data and information. In data analysis, we can quickly understand the market trend, order quantity, performance completion and so on through intuitive figures or trend charts, so as to intuitively absorb data information and contribute to the accuracy and real-time decision-making. For e-commerce websites, traffic is a very important indicator. In the above figure, we summarize the users (UV) and page views (PV) of the website into a unified data dashboard and update it in real time. Such data kanban, core figures and trends at a glance, which is clear to us. 2. Dimension decomposition When a single number or trend is too macro, we need to decompose the data through different dimensions to obtain more detailed data insight. When choosing a dimension, we need to carefully consider its influence on the analysis results. For example, when abnormal website traffic is detected, the problem can be found by splitting areas, access sources, devices, browsers and other dimensions. In Figure 7, the number of visits to the website on that day was significantly higher than that of last week. What is the reason? When we divide the traffic into dimensions according to the access source (Figure 9), it is not difficult to find that the number of direct access to the source has been greatly improved, which further focuses on the problem. 3. User Grouping It is a kind of user segmentation method we often say, which classifies users who meet certain behaviors or background information. We can also create a group of portraits of users by refining their specific information. For example, users who visit shopping websites and send addresses in Beijing can be classified as "Beijing" users. For the "Beijing" user group, we can further observe the frequency, category and time when they buy products, so as to create a portrait of this user group. In data analysis, we often carry out targeted user operation and product optimization for users with specific behaviors and backgrounds, and the effect will be more obvious. In the above picture, we select the users who failed to pay in a promotion through the user grouping function of GrowingIO, and then push the corresponding coupons. Such precise marketing promotion can greatly improve users' willingness to pay and sales amount. 4. Most commercial realization processes of conversion funnel can be summarized as funnel. Funnel analysis is one of the most commonly used data analysis methods, whether it is registration conversion funnel or e-commerce order funnel. Through funnel analysis, we can restore the path of user transformation from beginning to end and analyze the efficiency of each transformation node. Among them, we often pay attention to three points: first, from the beginning to the end, what is the overall conversion efficiency? Second, what is the conversion rate of each step? Third, which step loses the most, and what are the reasons? What are the characteristics of lost users? The above registration process is divided into three steps, and the overall conversion rate is 45.5%. In other words, 1000 users came to the registration page, and 455 of them successfully completed the registration. However, it is not difficult to find that the conversion rate of the second step is 56.8%, which is obviously lower than that of the first step and the third step. It can be inferred that there is a problem in the registration process of the second step. Obviously, the second step has the largest room for improvement, and the return on investment ratio is definitely not low; In order to improve the registration conversion rate, priority should be given to the second step. 5. Behavior trajectory pays attention to behavior trajectory in order to truly understand user behavior. Data indicators themselves are often just abstractions of the real situation. For example, if web analytics only looks at such indicators as user visits (UV) and page views (PV), it is absolutely impossible to fully understand how users use your products. Restoring the user's behavior trajectory through big data means helps the growth team to pay attention to the user's actual experience, find specific problems, design products and launch content according to the user's usage habits. The above picture shows a user's detailed behavior trajectory on an e-commerce website, from official website to the landing page, to the product details page, and finally back to the home page of official website. The conversion rate of website purchase is low, and the previous business data can't tell you the specific reasons; By analyzing the above user behavior trajectory, we can find some product and operation problems (such as the wrong goods, etc. ), so as to provide a basis for decision-making. 6. Retention analysis In the era when the demographic dividend is gradually fading, the cost of retaining an old user is far lower than the cost of acquiring a new user. Every product and service should focus on the retention of users to ensure that every customer is satisfied. We can understand the retention situation through data analysis, and we can also find ways to improve retention by analyzing the relationship between user behavior or behavior groups and return visits. In LinkedIn, the growth team found through data that if a new user comes in and adds more than five contacts (red line above), his/her retention rate on LinkedIn is much higher than those who don't add contacts (green line and purple line above). In this way, adding contacts is called one of the core means for LinkedIn to retain new users. In addition to paying attention to the overall user retention, the marketing team can pay attention to the user retention obtained from various channels or the return visit rate of registered users attracted by various contents, and the product team pays attention to the impact of each new function on the return visit of users. These are common retention analysis scenarios. 7.A/B testing A/B testing is used to compare the effects of different product designs/algorithms on the results. A/B testing is usually used to test the effects of different products or functional designs during the product launch process. The market and operation can complete the effect evaluation of different channels, contents and advertising ideas through A/B testing. For example, we designed two different forms of product interaction, and compared the visit duration and page views of the experimental group (Group A) and the control group (Group B) to evaluate which interaction form is better. A/B testing has two essential factors: first, there is enough time for testing; Second, the data volume and data density are high. Because when the product flow is not large enough, it is difficult to get statistical results when doing A/B test. A big company like LinkedIn can conduct thousands of A/B tests at the same time every day. Therefore, A/B testing is often used more accurately in the case of large company data, and statistical results can be obtained faster. 8. Mathematical Modeling When a business goal is related to various behaviors, portraits and other information, we usually use mathematical modeling and data mining to model and predict business results. As a SaaS enterprise, when we need to predict and judge customer churn, we can establish a churn model through user behavior data, company information, user portraits and other data. Use statistical methods to calculate some combinations and weights, so as to know which behaviors users conform to, and the possibility of churn will be higher. We often say that we can't grow without measurement, and data analysis plays a vital role in improving the business value of enterprises. Of course, it is far from enough to master simple theories, and practice makes true knowledge. We can try to use the method of data analysis in our daily work to analyze related projects. I believe that we can get twice the result with half the effort and create more commercial value. Third, as a product operation, data analysis tools must be used in the face of massive data. In the past, product operation generally went through four steps: data collection, data sorting, data perspective and data analysis to complete a data analysis. Usually use the following tools: data collection: data collation such as Python and GoogleAnalytics; data perspective such as Excel and SQL; data analysis such as Excel and PowerPoint; Excel and MATLAB. In this process, data collection, sorting and perspective will basically take up more than 70% of the time of product operation, resulting in the most valuable data analysis being squeezed. In order to complete the data analysis in a limited time, the data analysis value of product operation may be greatly reduced. GrowingIO product analysis chart Now, through GrowingIO product analysis, product operators can save a lot of time for data collection, collation and perspective. Simply integrating SDK, they can customize various visual charts (funnel analysis, trend analysis, distribution analysis, user portrait, etc.). ) Through GrowingIO, it really falls on data analysis and gives full play to the real value of product operation data analysis. As a product operation, it is also very important to use data tools well after having data analysis thinking and mastering data analysis methods. It can save us a lot of time and give full play to our true value. Zhiyiou Welfare: Click to receive GrowingIO product analysis (support website, App, applet) 15 days. Finally, young Xia, remember to like it when passing by ~