How to do data analysis when user growth occurs?

User growth is an unavoidable topic when working or looking for a job. This article shares the methods most companies use to increase user growth from two aspects.

01 Basic methods for user growth

1. Buy volume

In Internet companies, buying traffic accounts for the majority.

Generally divided into:

(1) Search engine advertising , including app markets, short video platforms, and search platforms of Tencent, ByteDance, and Baidu.

(2) Secondary channel advertising , including: advertising within free apps (such as reading novels for free by watching ads), advertising within games (such as), advertising on long video websites, etc.

(3) Ground promotion . Yes, you read that right. Ground promotion can be large or small. Large-scale ground promotion is like subway ads and billboards in squares, while small-scale ground promotion is like handing out flyers and collaborating with certain brands.

General Internet companies mainly rely on search engines and secondary channels. Only slightly larger venues, such as Xiaohongshu, Zhihu, and Touteng, will have ground promotion teams.

As for the Internetization of traditional enterprises, the methods are similar, but the structures are different. Generally, enterprises with chain stores will mainly rely on ground promotion because of the low cost, quick results, and less search and channel purchases.

Therefore, the analysis methods for user growth are different in different industries, which we will discuss below.

2. Fission (referral)

There is a scenario that everyone is familiar with using this method - Pinduoduo cuts prices.

In fact, this method looks very simple and crude, but there are many details that need to be paid attention to.

There are also several fixed ways of fission:

• Activity fission: By designing activities and matching incentives, users can actively bring in other people to get discounts. This is the most common method used by PDD. Taobao and JD group purchase orders also belong to this type of fission.
• Product fission: Through the design of some product mechanisms, users are encouraged to actively share at certain points. For example, sharing will get you 3 days of VIP membership, or sharing good skins and cards drawn in the game, which are generally permanent.
• Private domain fission: This method is commonly known as "old customers bring new customers" and is commonly used by traditional high-customer-order enterprises, such as education and training, medical beauty, fitness, etc. Old customers bring new customers. Some large-scale ones will turn members into agents and make large-scale rebates, turning it into a distribution model.

In most Internet companies, there is no dedicated department to do activity fission and private domain fission. Product fission is the most common, because it only requires designing rules and product guidance, and the cost is relatively low. But at the same time, the effect will be relatively poor.

3. Optimize the process

The content of this section is generally used in conjunction with AB experiments, and most of the time it is used to resolve disputes.

for example:

Users come in, but the conversion rate is not good. Is it a product problem or a problem of buying volume? !

The product fission is not working. Is it because the entrance is too deep, or the product content is not good and users are unwilling to share? !

A small part of the time is spent on tuning:

• Multiple activity plans, select appropriate parameters to achieve the best fission effect.
• Conduct horse racing experiments on multiple product locations and product solutions to select the optimal path.
• Experiment with different incentive schemes in multiple different regions to minimize costs.

Companies that can do such detailed work as process optimization are generally of some size. Although AB experiments are mostly used to pass the buck and fight for credit, they will definitely be tested in detail during the interview, so you can pay attention to it.

In addition, when doing serious optimization, the principles of AB experiments are generally not very important because they have been productized. However, "how to do multiple experiments at one time" becomes more important. Therefore, if there are multiple AB groups in parallel in the business, understanding the bucketing rules and principles of AB experiments will make you more competitive.

02 Analysis of user growth

1. How to analyze the work of buying traffic?

Take some examples of work content:

(1) Each app market has different rules for buying traffic. For example, the vivo market has banner ads, which can be charged based on exposure. The IOS market generally relies on keyword optimization and ranking.

(2) Or using Toutiao content delivery is different from SEO/SEM delivery. Generally, content platforms have recommendation systems. If you can accurately define valuable users and then send them back to the platform, you can effectively debug the algorithm to improve the ROI of the advertising plan.

(3) The data that can be obtained by each promotion platform is different. For example, the domestic big data and basic API can only obtain aggregate data such as cost and revenue, while the foreign appflyer can obtain user-level attribution data. The difference in data granularity will also lead to different levels of analysis and costs.

There are rigid evaluation indicators for buying traffic: LTV and ROI. These two indicators are the key to evaluating whether the delivery is good or not. Therefore, all the work related to buying traffic can be carried out based on these two indicators.

Next, we can disassemble the process indicators:

(1) Conduct funnel analysis based on product process indicators such as click-through rate, bounce rate, and page conversion rate.

(2) Basic user attributes such as source channel, IP address, registration period, etc., pay attention to the MECE principle when breaking down.

The above analysis is relatively formulaic because it is relatively detached from business logic.

For a more in-depth analysis, you will need to use the buying knowledge of the different platforms mentioned above to judge the pros and cons of the promotion plan and the quality of the materials. Due to space reasons, I will not list them one by one. You can learn about the characteristics of the platform used by your company on your own, and discuss it in detail if you have any needs.

2. How to analyze user fission?

The analysis of user fission is also related to the specific working method:

• The fission of activities is closely related to the business model. Generally, what needs to be analyzed is the parameters of the activity, the target group of the activity, the form of the activity, the content of the incentive, etc. to find the optimal solution.
• Those who engage in product fission generally break down productized content such as penetration rate, sharing rate, and click-through rate of shared links .
• Those who engage in private domain fission will pay more attention to whether the incentive and mechanism of cashback are reasonable and whether there is an anti-pickup mechanism. The second is to analyze the agents and conduct hierarchical management of multi-level agents.

Similarly, there are various types of user fission, and there is also a core indicator for analysis: the fission factor.

The formula for the so-called fission factor is:

K factor = (number of invitations each user sends to his friends) * (conversion rate of people receiving invitations to new users)

When K>1, the user base will grow like a snowball. If K<1, the user base will stop growing through self-propagation when it reaches a certain size.

However, in order to accurately calculate the K factor, it is very important to understand the user's bloodline and calculate the user's parents and children.

The so-called user lineage means that we need to know that A introduced B, B introduced C, and we need to know that A is the superior of C. Generally, at least three levels of lineage are required to calculate the fission factor.

3. How to analyze process optimization?

The process optimization here basically refers to experimental analysis.

The experimental analysis is divided into two cases:

(1) In scenarios where experiments are used to assign credit or pass the buck, verify whether the preset conditions, traffic, and execution process meet the experimental requirements. If they meet the requirements, use a T test or chi-square test to get the results. If they do not meet the requirements, do not give the experimental results and just give feedback on the experimental requirements.

(2) For serious experimental optimization, the main purpose is to verify whether the traffic distribution meets the requirements and manage the experimental process. Subsequent experimental conclusions and calculations can be drawn according to the standardized AB process. This situation is actually simpler.

Therefore, for this part, basic knowledge is actually the key. The theoretical basis of AB experiments and the technical basis and understanding of multiple experiments in parallel will ultimately affect the experimental results.

03 Conclusion

The ultimate business logic of user growth changes with market changes.

For example, from 2013 to 2018, the story of mobile Internet made sense, and everyone believed in the story of charging 1 yuan each to 1.3 billion people.

Therefore, the growth in that era generally referred to DAU growth, and it didn’t matter whether there was payment or not.

Because in that era, the business logic of Internet companies was to sell a story to investors to raise funds, and then sell the company to cash in.

After many years of iterations and many stories, user growth is now more responsible for conversions and profits.

In this scenario, we will be more pragmatic in analyzing the relationship between users, products and channels and move towards normal business logic.

Author: Wang Hao; WeChat public account: Wang Sir who only speaks human language