Former Facebook Engineer Qin Chao丨Facebook Growth Secrets and Practical Methodology

Last weekend, I attended a salon event called "Things You Don't Know About User Growth" co-organized by Xtecher, DeepShare, and Peking University Incubator. I was deeply touched after listening to it. On the one hand, practical cases from Tencent Weibo and Tencent Weishi in the past continued to emerge, which coincided with what the sharing guests said; on the other hand, in Tencent's other product lines, there are also a large number of growth cases in internal OA. I didn't organize them well before, but now I suddenly realized and decided to continue to dig deep into some internal cases to enrich this growth system.

I got the authorization from Xtecher to share the speech transcript of the day with you. Let’s start with the sharing of FreeS, a former Facebook engineer and technical partner and Zhihu celebrity Qin Chao:

Today I will share my experience in Growth Hacking at Facebook. Why should you learn Facebook’s Growth Hacking technology?

When I joined the company, I started working on the Facebook Phone project. It included hardware, OS, SDK and App, but the project did not go smoothly and was cancelled.

When starting a business, people tend to be more pessimistic about the company. For example, many people asked me if I felt pessimistic at Facebook. In fact, I was in the company for four years, and I only felt good for more than half a year. The rest of the time I felt that the company was going to fail. The market value was also falling, and many major projects failed. Google also launched many good products. Therefore, when starting a business , what is more important is to know how to persist.

Why should you learn Facebook's Growth Hacking techniques?

When I joined the company, Facebook's monthly active users were 550 million. When I left the company, the number reached more than 1.6 billion. Now it is nearly 1.7 billion. Daily Active Users increased from 250 million to 950 million, and mobile users increased from 20% to 85%. Monthly active users and daily active users are growing rapidly on mobile devices. So how does it do that?

If you look at the user growth curve over the past four or five years, you will find that it is like a smooth straight line going upwards. But if you look at the user growth curve from a month or a week, you will find that it is very steep. What is the reason for this? We have a team that makes a lot of adjustments every week to see how the data changes, find out the reasons, and take corresponding measures. If you do well, keep the good experience.

Facebook has a strong Hack culture, which is: get started right away, get it done quickly, and iterate continuously.

One thing that is very taboo on Facebook is that after an idea is proposed, many people come together to discuss it, but after two weeks there is no progress. This is absolutely taboo in the company. Once an idea comes up, they will make a prototype, test the effect, and then develop version 1.0 in a month and push it to users to test the feedback. Such characteristics are non-traditional, technology-driven, and data-centric.

Boss Zuck is very focused. He focuses on growth and has always emphasized user growth within the company, encouraging everyone to use different ideas to achieve user growth.

Back to the Growth hacking technique itself, what I want to emphasize is its difference from traditional growth. Many American companies often spend a lot of money on advertising, purchasing channels, and various marketing activities .

This is a typical user usage diagram of an Internet product, which shows that users will be screened through layers of funnels. When you first acquire a user, you are attracting new users. How many people are activated after attracting new users and how many people stay on your product? These people can create value. Later, if users like it, they will share it. This is the typical AARRR model.

The real difference between it and traditional marketing is:

The traditional market is to spend a lot of money to let users know about the brand and attract users to use your product, but in the end there is no good data to track how many users start paying, what the retention rate is, and how many people are sharing. In this regard, it is difficult for traditional enterprises to achieve this because there is no connection between the market and technology. Growth hacking is essentially the result of collaboration among a technical team, an operations team, and a product team, so it can penetrate into every detail to obtain relevant data.

The biggest feature of Facebook's Growth hacking is that it is led by Zack himself. This is what I want to emphasize. If a company wants to establish Growth hacking, your CTO must have his own ideas, and he himself is the promoter of Growth hacking.

People often ask me how to build their own growth team. I personally think that the first thing is that the CEO must understand growth hacking and support it. The second is to have support in products, technology and operations. The worst thing is that you set up a growth department, but there is a lot of interference, which makes it difficult to move forward. Third, if it is a startup company, then your entire company is a Growth.

Here are 4 steps to apply Growth Hack technology on Facebook:

1. Create a data panel from raw data

Pay attention to daily activity, monthly activity, duration of stay, user retention and other data

As for user stickiness, Facebook will focus on an indicator: L6/7, which refers to the proportion of users who log in for 6 consecutive days within 7 days. It can sense the user's turnover rate. For example, if your daily active users are 50,000 today and 50,000 tomorrow, you can't tell the difference. This indicator records the loyalty of a single user on the platform.

Others have different indicators for specific purposes, such as adjusting by measuring Likes and comments, and adjusting based on core business indicators.

2. Analyze and study the data panel

Understand the core business of the product

Monitoring + Alerting

Learn to use data to reversely promote product function iteration

So how can we use data to drive product functionality iteration? Its essence is to find patterns from complex data. There are 5 steps:

Grouping

Compare

Find the difference

Assumptions

Argument

For example, how did Facebook’s “people you may know” feature come about? This feature is now called "People You May Be Interested In". It was first launched by LinkedIn. When Facebook launched this feature, the number of users exceeded MySpace. It is one of Facebook's most effective growth features.

At that time, we found that some users were very active and some were very inactive. We grouped these users and compared them, identified their characteristics, and found the differences in the average number of friends and the completeness of their personal information between the two groups. So we make one of the assumptions: the more friends a user has, the more active he is. Then we develop and test this function for 1% of users. A month later, we observed these users and found that this hypothesis was confirmed. The activity of users who activated this recommendation function did increase, so we developed this function for all users. Even today, after Facebook has gone through many revisions, its “people you may know” still exists.

Another one is the degree of completeness of personal information. Facebook made such a small tool back then, and many domestic social networks also made it, such as Maimai. The more complete your personal information is and the higher your activity level is, the higher the user activity will also be.

Data reverse deduction is very important, but many companies do not pay attention to it. For example, Zhihu.

Zhihu has also used this method. Zhihu initially imitated Quar, which is a Facebook-based platform. Zhihu did something. It surveyed very active users on Zhihu and a group of inactive users, and then made a comparison. They found that there was a difference in the number of questions answered between the two. After analysis, they finally came up with a hypothesis that users who answered 3 or more questions would generally stay. So the entire operation staff no longer attracted new users, but disguised themselves as users and constantly attracted other users to answer questions, which gradually improved the retention rate.

For different products, there are specific values ​​to improve retention rate. For example, after adding 10 friends on Facebook, following 30 people on Twitter, and sending 2,000 messages on Slack, the user can be converted into a paying user. For different products, you need to find your own specific numbers to stabilize users.

3. Growth Funnel

Attract new customers: from channels, advertising, activities, ground promotion , etc.

Observation to improve retention: good products, monitoring, and recall mechanisms

Encourage users to share again: there is a share button on the product details page

The sharing function is very simple. Now when you go to a product details page, there will be an outward button in the upper right corner, which makes it easy to share to Moments or Weibo. Although Growth Hacking is very popular, the essence of user retention is to meet user needs, which is an internal factor, and Growth Hacking is an external factor. External factors require internal factors to work, so when you make a product, you must understand what the core needs of users are. Only when this is met, Growth Hacking will be meaningful.

4. Grayscale release and A/B testing

Plan: Develop a revision plan based on new features

Expectations: How the data will change

Set up multiple versions: gradually open to users

Clear: Clear old versions

The essence of this is very simple. After a function is developed, it is not directly available to all users. At Facebook, it is first used internally by employees and then released to 1% or 5% of users. See how this part of user data changes. At least it should not get worse. If you think it is good, then open it to users step by step. The above points are relatively theoretical. Let me briefly explain to you that one thing Facebook needs to do when developing a feature is to plan how the data will change, then give an expectation, gradually open it to 100% of users, and finally embed the code.

Let me give you an example of the redesign of the Facebook homepage. This was a very failed example within Facebook. Let me tell you how we used data-driven methods to do this.

This is the homepage version from 2009. At that time, a group within Facebook thought of many ways to improve the homepage. For example, when Pinterest came out, they wanted to try the waterfall flow. We tried several versions, and finally decided on this version, which has large pictures and a flat style. The navigation bar on the left is represented by icons. The entire development took about half a year. After it was completed, everyone thought it was very beautiful, and the employees found no problems after trying it, so we developed it to 2% of users, but problems arose.

(The improved new version at that time)

Users' online time began to drop, and their activity levels also dropped. As the online time dropped, advertising was also affected. When we went back to check, we found that there was a bug when they first collected employee data and there was a problem when they were setting the data points. After the modifications, we found that the data had already dropped when they first started counting. The members of this group began to think that it might take time for users to adapt to the new version. At the same time, the user sample of this survey was too small, so it might be that this group of users did not like it. So they decided to observe for a while and made a lot of optimizations, but user activity continued to decline. The new version was used by 12% of users, but it ended up falling by 18%. This situation lasted for about three months, and the data was still not good. The top management did not approve this version and reverted to the previous version.

Looking back at the previous versions, they are a bit like the hao123 version, or the version from the 80s. Watching the new version, it looks much better. So where is the problem? We analyzed it later. Why is this version of data bad? We found a group of users to try out the system. Some of them left the page after clicking on it for two or three minutes. We asked them why their online time was so short. They said, “Everything is good and beautiful, but the information can be read very quickly.” However, the old version had many entrances to information and people could read the information for a long time.

So all of this requires collecting data, then analyzing the data, and making choices based on your strategic needs. However, the leader of this team did not do their homework and did not collect data or use data for verification in the early stages of the project, which resulted in such a large team wasting 6 months.

Finally, let me summarize four points. For Chinese companies, when using Growth hacking, please note:

（1）Learn to speak with data-driven thinking

(2) To prevent dictatorship, leaders must support

(3) Relieve server pressure

(4) The last line of defense for major bugs is to shut down the server

In short, Growth hacking is an addition, not a cornerstone. It is a change from 1 to n, not a change from 0 to 1. Listen carefully to user needs, but don’t act according to them. Extract the essential needs of users to create products. Finally, I would like to emphasize that if you are a startup company, the four steps of Growth hacking may not all apply to you. It is better to first have a group of seed users and then switch to Growth hacking.

APP Top Promotion (www.opp2.com) is the top mobile APP promotion platform in China, focusing on mobile APP promotion operation methods, experience and skills, channel ASO optimization ranking, and sharing APP marketing information. Welcome to follow the official WeChat public account: appganhuo

[Scan the top APP promotion WeChat QR code to get more dry goods and explosive materials]

This article was compiled and published by (APP Top Promotion) by @我这一生. Reprinting this article requires the consent of Top Promotion , and please attach the link to this article!