If you work in the Internet industry, do you really understand traffic analysis?

In my current job, I come into contact with many friends who do BI analysis in Internet companies. I would like to take this opportunity to organize what I have seen and learned and share it with everyone. At the same time, I also hope to increase communication with friends who share the same interests through sharing. We welcome your comments on any deficiencies or errors in the article.

There are already a lot of articles and books about data management, and I am just sorting them out here. Data analysis is also for the development of the company, or to put it bluntly, for the company's profitability and sustainable profitability. From this perspective, let’s analyze one by one what data needs to be analyzed in the Internet industry, how to analyze it, and what is the value of the analysis. I will divide it into four parts: income-related data analysis, cost-related data analysis, risk (for sustainable development)-related data analysis, and comprehensive management.

The following will introduce them one by one (updated in stages).

Chapter 1: Income-related data analysis

The business model of the Internet is ever-changing, but its profit model can currently be roughly divided into the following three types: one is to sell goods or services to users, among which e-commerce and o2o belong to this model; the second is to make profits through advertising, typical examples of which are Google, Baidu and other platform-based Internet companies; the third is to charge users directly, and most game companies currently belong to this model. Different income models also have different data indicators, and we will introduce them one by one.

1. Selling goods and services to users

E-commerce companies and o2o companies mainly make profits through this model. The company's revenue is accumulated from orders, and its revenue status can be reflected through the order status. Orders are generated when users purchase related goods or services. It can be said that users and goods or services are the two basic elements of an order. The company's revenue decline, growth, and anomalies can ultimately be traced back to these two elements: users and goods. In this way, we break down the company's revenue-related data into three major modules: users, products or services, and orders.

user

The company's revenue and orders are all generated by user consumption. The user's consumption process can be divided into the following four stages: traffic, conversion, consumption, and retention. The ideal situation we hope for is that a large number of users come in and generate consumption, and continue to generate consumption. However, reality is usually far from what we hope for. What we can do is to analyze these data and adjust strategies based on the data to bring reality closer to the ideal situation.

We generally divide users into new users and old users, as shown in the following figure:

Regardless of new or old users, we are concerned about two things: one is traffic generation (attracting new users), and the other is conversion, which are ultimately reflected in the form of data, namely traffic and conversion rate .

drainage

A shopping mall built in the wilderness is useless if no one comes in, no matter how luxuriously decorated it is. According to CNNIC statistics, the number of Internet users in China reached 688 million in 2015, with a stable growth rate still maintained at 5.7%. JD.com's traffic in the fourth quarter of 2015 reached 131.9 million, up from 82.8 million in the fourth quarter of 2014, a year-on-year growth rate of 59%. These figures tell us that resources are scarce, but there is always room for growth. We need to be prudent and achieve the maximum value of each channel and each type of traffic source.

Analysis goal: Ensure the stability of traffic by analyzing it, and try to increase it through adjustments.

Analysis angle:

1. Observe traffic patterns to facilitate event arrangements and service adjustments
2. Detect traffic anomalies, analyze the causes and make timely adjustments
3. Observe the traffic structure, analyze its rationality, and make adjustments
4. Track traffic, measure campaigns or adjust results

Analytical methods:

• Let's first understand some basic data indicators about traffic: the number of visitors (uv), page views (pv), and the number of visits (visits), which are commonly used data indicators to measure the amount of traffic;
• Average visit depth (number of views/number of visits), average dwell time (total dwell time/total number of views), and bounce rate (number of bounces/number of visits) are indicators used to measure the quality of traffic.

There are many ways to achieve the above goals. Visualizing the data and observing the data from a reasonable angle will make the data presentation clearer and reduce the difficulty of discovering problems. The following will implement data analysis from various angles in the form of charts.

1. Observe traffic patterns to facilitate activity arrangements and service adjustments

From the above figure, we can find the following patterns: during the day, visits are concentrated in the working hours from 9:00 to 11:00 and from 14:00 to 17:00. During the year, the number of visits is relatively large around the Spring Festival, and visits are concentrated on weekdays. The rules of most Internet businesses are different from the above situation. Generally, 2c businesses will have huge traffic during rest time, which may happen to be the opposite of the above situation, but it does not hinder the above analysis method.

Generally speaking, traffic is distributed according to daily time periods, seasons, holidays, and weeks. Therefore, you can put the above aspects together on the same page for observation, so that you can fully understand the access rules of the application. And by selecting channels and businesses, the access patterns of specific channels and businesses can be observed.

Analyzing the patterns of traffic has a significant impact on activity results and business adjustments. For example: Company A wants to launch a two-day, two-hour event each day where customers can get a 50 yuan discount on purchases of 200 yuan or more and receive a free soap. Then the event should start on Wednesday or Thursday at 9 or 10 o'clock in the morning. Only in this way can it be known by most users within a certain period of time. After all, the advertising time cost of the event is money. In addition, the company's business layout can be adjusted and reasonable cost control can be carried out according to the intensity of visits in different periods.

2. Detect traffic anomalies, analyze the causes and make timely adjustments

By observing the above figure, we can find two abnormal phenomena:

1. Traffic is distributed according to a weekly pattern, with higher traffic on weekdays and lower traffic on weekends. However, in the above figure, May 2 and April 1 are Mondays, and the traffic is also very low. Observing the calendar, we find that these two days are the May Day and Qingming holidays, which are still rest days, so the traffic is not high. This is normal.
2. From the traffic graph from March 21 to April 17, the traffic on weekdays generally remained at around 2,400. However, observing the graph from April 18 to May 15, it was found that the traffic began to decline from April 19 and rarely exceeded 2,000, which means that the traffic has declined significantly in the past month. The reasons may be that the opponent purchased the bidding ranking, your own SEO was not done well, etc. Once a problem is discovered, the specific causes must be further analyzed based on actual conditions.

Generally speaking, traffic is measured in weeks, and there are many cases of periodic distribution. It is easier to find problems by extending the perspective and looking at data for several weeks at a time. Comparing data over a period of time with historical data can also help identify problems.

In addition to the simple monitoring of traffic anomalies in the above figure, the traffic can be further decomposed, as shown in the figure below. By linking charts, the traffic conditions of specific channels or businesses can be observed to track and locate the problem. For example, through further analysis, it was found that the traffic decline that began in mid-April mainly occurred on the PC side, so the scope of the problem can be further narrowed. Facilitates problem solving.

3. Observe the traffic structure, analyze its rationality, and make adjustments

The traffic structure can generally be divided into channel structure, business structure and regional structure. Understand the traffic composition by querying the proportion of each structure over a period of time.

As shown in the figure above, in the channel, the proportion of PC is relatively too large, while the proportion of app is not high. App has greater stickiness for users, so we should analyze the reasons why the proportion of app is too low and find ways to increase the proportion of app traffic. The line chart below can track the traffic situation of each channel, analyze whether the unreasonable proportion is short-term or long-term, and assist in the analysis of the problem.

What proportion is reasonable varies in different scenarios, but generally speaking, the proportion of paid traffic should not be too high, and the proportion of general channels should dominate. It is even more different for each business. However, the rationality can be analyzed by comparing industry data or competitor data, of course, the premise is that relevant data can be obtained.

Usually there are many sources of channels, including independent visits, search engines, Taobao payment, JD payment, etc. Some people will analyze the quality of each channel based on the proportion of channel traffic. Measuring quality based solely on traffic is not comprehensive; it needs to be combined with conversion rate and ROI. The details will be considered together when writing about conversion rate later.

4. Track traffic, measure activities or adjust results

Traffic tracking generally means monitoring traffic, observing changes before, during, and after an activity, and evaluating the effectiveness of the activity. Generally speaking, traffic will increase significantly during the event and will decline somewhat after the event, which means it is a successful event. If the traffic does not increase much during the event, or if the traffic drops sharply after the event, or even if the traffic is much lower than the normal traffic before the event, it cannot be considered a successful event.

Of course, if you want to analyze the effectiveness of an activity, you need to track more than just traffic. You also need to track conversion rate, number of orders, and transaction amount. A targeted introduction will be given later.

In addition to activities, companies may often adjust channel investment, page layout, functional improvements, etc. After each adjustment, the traffic is tracked and observed to analyze the effect of the adjustment. Here we only introduce the tracking of traffic. After optimizing and adjusting products or channels, you also need to track conversion rates, etc. The conversion rates will be introduced later.