The whole process of obtaining 100,000 seed users from 0

Written at the beginning: Product Operation from 0 to 1 (Practical Edition): A series of articles on how to acquire seed users , mainly focusing on six aspects: product, crowd, channel, copywriting, activities, and optimization.

As for why I wrote the practical part first:

1. The practical part is difficult and tiring to write. It is hard to write it without real leadership and experience.

2. Most of the articles on the Internet talk about theory and are separated from the practical steps. Generally, newbies can understand them clearly but don’t know how to put them into practice.

3. From practical operation to theory, better understand the theoretical level and help novices grow quickly.

Entering the mainstream : I have been very lucky recently. I have experienced the cold start phase of two tool products. These two products have begun to grow. I personally feel it is necessary to summarize my thoughts and pitfalls in this regard.

When I took over the cold start of the product, I read a lot of articles online. Many of them analyzed from an ideological level or gave general direction guidance, but there were problems in the actual implementation. In other words, what they said seemed right, but when they were put into practice, they were all wrong. In fact, some people who write articles simply summarize other people’s thoughts and share them without even understanding the basic operating rules. There is really no value in reading this kind of summary.

This article is different from other articles on the market that talk about theories and cases. It will share from the perspective of implementation. It is a routine of first having products and then operating them. It is slightly different from the operation method of first operating and then having products. Another product belongs to the operation routine of first operating and then having products. (If I have time, I will share from 0 to 1, the rapid implementation of AI products, and write from the idea of ​​having users first and then products).

This article is more suitable for newbies who have just entered operations. Most of the operations are at the implementation level. There are some ideas, but they are not important summaries. The real ideological level things will be summarized in the ideological section. This operation is how to grow the product from 0 to tens of thousands of users. People in the industry already know that the user target is 100,000.

When doing something, you must understand the purpose or meaning of doing it, otherwise it would be blind. For product operations, there will be North Star indicators at every stage.

How are the North Star indicators of a product determined? Generally, they are issued by the leader or reported by the operations department and approved by the leader.

1. Objectives

The first phase of the North Star Indicator: quickly acquire 10,000 users and verify whether the product has achieved product/market fit. The boss's words have two important instructions:

1. Quickly acquire 10,000 users. Quickly is too general and you can only communicate with your leader before this.

2. Whether the product has reached PMF. PMF should also have an indicator, which state is considered to have reached PMF.

So call the boss, product and operations, and preferably the technical staff, to discuss together how fast fast is and what state is product/market fit. General indicators must be quantified, and indicators based on feelings are not reliable.

The final definition is: the fast time is 1 month, and the pmf status is the next-day retention rate is greater than 20%.

So the North Star Indicator was changed to: acquire 10,000 users within 1 month and verify whether the product’s next-day retention rate is greater than 20%.

The operation goal is very clear: to acquire 10,000 users within 1 month, and the product goal is also clear: the next-day retention rate is greater than 20%.

All indicators must be quantifiable in order to have the next operational plan.

2. Product Positioning

With a goal in mind, we need to position the product. When I came into contact with the product, it had already been produced but had not yet been launched. For a startup company, optimizing product details before the product is launched is a fatal thing.

In any case, the product is already there, so let’s analyze it from a product perspective, position the product, and get started!

2.1 Product attributes

The product is the aixcoder intelligent programming plug-in, a product that only programmers will use. The vast majority of programmers write code on computers, which determines the product's usage scenario: programmers sit in front of computers and program.

Although they are all sitting in front of the computer, their operating systems, programming tools, and programming languages ​​are different, and the product usage scenarios will also be divided based on these three dimensions.

2.1.1 Operating System

Windows, Mac, Linux

2.1.2 Programming Tools

IntelliJ IDEA, Eclipse, PyCharm, Android Studio, PhpStorm, Visual Studio Code, sublime...

2.1.3 Programming Language

Java, python, php, c++, js, go...

According to the product usage scenarios, there are so many operating systems, programming tools and languages. Which one is the main focus? We conduct data statistics and judgments from three aspects.

2.2 Product Data

2.2.1 Questionnaire

Where should I send the questionnaire? Often, we see people sending links to questionnaires on WeChat Moments. The credibility of the data obtained in this way needs to be examined. After all, a friend may fill it out for you after seeing it, but he is not someone in this industry, so the data he gets is not credible.

The best way to distribute questionnaires is to join several related QQ groups. The reason why it is recommended to start with QQ groups is because WeChat groups are not easy to find. QQ has a search function that can find related QQ groups.

Sending a link to a questionnaire and then sending a red envelope has two benefits:

(1) Avoid being kicked out by the group owner or administrator

(2) Mobilize the enthusiasm of users in the group

From an operational point of view, it can avoid being kicked out of the group by the administrator, and can collect feedback from 10% of group users. Without sending red envelopes, more than 3% of user feedback can be collected.

2.2.2 Internal Discussion

When choosing the main channel, the boss and product manager usually already have an idea, but their ideas are sometimes not reliable, and data must be produced to decide which channel to choose; or the product manager may have already compiled statistics on the product data when making the product, and the main direction can be chosen based on the product manager's data.

2.2.3 Third-party data

Every year, there will be statistics agencies that release some data, and these data can be obtained through the Internet. However, it is not recommended to search through Baidu. A lot of the data searched out are unreliable. If you don’t know whether there is similar data, it is best to ask a professional.

Regarding programming languages : It is still easy to obtain the usage of programming languages ​​in the programming industry.

Programming language rankings in March 2019

Regarding usage scenarios : Windows, Mac and Linux are all supported, so there is no question of which operating system to launch first.

Regarding programming tools: This requires a survey of users. Based on the user survey data, it is found that some people use editors that are not supported by the product. Analyzing from other data, it is impossible to promote the product according to the editor unless it is put on the official market.

So we arranged the delivery based on the user acquisition scenarios of the product, starting with language: follow the language ranking from top to bottom, and deliver according to the languages ​​that the product already supports. The operating system and programming tools are not considered for now.

3. Crowd Positioning

In fact, this product has already determined the target group: programmers. Regarding the classification of programmers, the classification of tool products cannot be based on universal products, but on gender, age, and preferences. Here they can only be divided into junior programmers and senior programmers.

3.1 User attributes

You can also use third-party tools to perform exploratory segmentation, such as Java by region, age, and interest distribution.

From here, we can conclude that our user group is: men aged 30-39, because their interest distribution is similar.

The conclusion drawn here is superficial and an extremely irresponsible inference. Because this programming tool is suitable for any programmer, our user group is all programmers.

3.2 User Hobbies

The access device is divided according to the data of Java users searching and accessing Java. (ps: the access device here is different from the product usage device)

According to the users' daily access data, we found an interesting phenomenon: the data on Saturdays and Sundays were relatively low, mainly because programmers did not go to work on Saturdays and Sundays. So we gave up the idea of ​​promoting products on Saturdays and Sundays. (The product usage data later also proved our idea.)

3.3 User gathering places

Where are the users? Does the so-called fish pond really exist? Actually, there are. For example, a QQ group is a small fish pond.

The largest gathering place for programmers in China: CSDN and 51CTO.

Since there is a fish pond, is there any way to fish from it?

The common method that comes to mind is posting, but this method is too slow and cannot meet the company's goals. As for how to play these channels, an introduction will be made within the channels.

From the crowd positioning, we can conclude that our product is targeted at all programmers on the PC side, and the time is working hours from Monday to Friday.

4. Channels

I am very happy. Channels are my strong point. Since I come from online marketing, I have worked in SEO, SEM, self-media, e-commerce, social media, and EDM.

According to my personal understanding, every channel has traffic, it is just a matter of how much traffic it is and whether it can support the company's business goals.

4.1 Channel Division

According to the three types of customer acquisition channels, we divide the channels into: word-of-mouth channels, organic channels and paid channels.

Word-of-mouth channels include: social media (such as WeChat, Weibo), friend recommendations, community recommendations, etc.

Organic channels: SEO, content marketing, EDM, submissions, etc.

Paid channels: SEM, Guangdiantong, Fanstong, sponsorship, etc.

In fact, every operations staff will think of these channels, but many people (including operations directors) do not understand the characteristics of the channels. When making a plan, they can only talk nonsense and come up with a plan that cannot be tested or implemented at all.

Sometimes a product is a good product, but it is ruined by people who don’t understand operation and promotion.

If you are lucky enough to work in a large company, this problem of channel division almost does not exist. Traffic from any channel can be enough to make your product a six-figure star launch product. However, most operators are not so lucky and can only attract users according to these channels.

4.2 Channel Testing

With so many channels listed out, as for which one to launch first, we need to evaluate it from several aspects, including situation, cost, time investment, output time, and scale.

Confirm the mastery: Degree of understanding of the channel, who in the team knows the channel better, let the person who knows the channel better confirm whether the channel can be used, calculate it based on 5 points, see if our channel mastery score is qualified, and calculate other inputs, and discuss internally whether the channel can be tested.

Here are just some simple lists. In fact, when you are considering and calculating, you should be more detailed than this table. As for why so many people are reached in a day, why so much money is spent, and how many people are finally converted, you must list them with reason and evidence, otherwise the leader will question it or fail the test.

When faced with other people's questions, channel personnel must be able to explain in detail and stick to their own opinions. When disputes arose, many people believed that EDM had no effect in China, and everyone agreed that the EDM channel should be abandoned.

For me personally, I know how to do it effectively and insist on doing EDM. In the last day, I sent 10,000 emails, which converted about 500 download users. The conversion rate was about 5%. The cost of one email was about 0.3 cents, which means the cost of one user was 0.6 yuan, a very cost-effective ROI.

The channels selected in the first round are: EDM, SEM, sponsorship, and paid submissions

The test order is: EDM, SEM, sponsorship, paid contribution (some people may ask why paid contribution is launched again, because there are too few channels and no room for expansion)

4.3 Channel segmentation:

4.3.1. EDM channels are divided according to languages ​​and IDEs. The languages ​​include Java, Python, PHP, JS, C, and C++. According to IDEs, the number of users is too small, so we will give up for the time being.

SEM: For the time being, the only channel is Baidu. Considering that programmers only use computers and give up on mobile terminals, the only channels are Baidu & PC.

4.3.2. Sponsorship: Sponsorship is provided according to comprehensive fields and vertical fields. Comprehensive fields include csdn, 51cto, etc., and vertical fields include java and python.

Of course, after the segmentation here, there should be another table. I will use EDM as an example to show this table because some data are indeed inconvenient to disclose.

After segmenting the channels, preparing what is necessary, going online for testing, and finally getting the test results, we estimate that the unit price for acquiring a user is around 15 yuan.

Test results: The test results are calculated based on segmented channels and the KPI of the overall goal. The assessment indicator is the unit price of registered users.

Baidu & PC: 150 yuan, sponsorship: csdn: 150, 51cto: 60; vertical field java1: 6 yuan; EDM: java: 0.6, python: 1.5, spring: 0.9.

4.4. Channel life:

Each fixed user channel has a certain lifespan. As investment in a single channel increases, the channel's traffic source will dry up. This does not include distribution channels such as DSP, SEM, and SEO.

4.4.1、Fixed user channels:

Because of the fixed user channels, the general user group changes very little. You acquire users from here, but there is only outflow and no inflow of users. Finally, the users in the pool are gradually exhausted, and you will eventually have to change channels.

4.4.2、Flow channels:

The channels for mobile users generally have an inflow and outflow of customers. The users in the pool will never be exhausted, and the inflow of users is far greater than the outflow of users. Such a pool can be used continuously.

After the channel testing is completed, you can screen out suitable channels and increase delivery, and suspend delivery to unsuitable channels. Problems that we were not expecting still came. Due to rule changes, the EDM channel was completely obsolete after two months and had to be replaced by other channels.

That is to say, when doing channel testing, there should be at least 1 main channel and 2 auxiliary channels. Once there is a problem with the main channel, increase investment in the auxiliary channels in a timely manner to avoid being in a panic after the main channel has a problem.

As for whether the copy should be written before or after the channel, now I think it should definitely be written before the channel is released, but what is certain is that only after the channel and the target audience are determined can we determine the writing style of the copy and what it will look like, so it is better to write the copy after the channel.

Of course, if you have different opinions, you can raise them and we can discuss them together.

5. Copywriting

The copy here includes the text on the advertising page, the text on the website, and soft articles, and does not only refer to the text written in the advertisement and the short text on the website.

As for how the copy is written, it still depends on the role of the copy in the entire process.

5.1 Copywriting

Copywriting is inseparable from channels, products, groups and the role in the channel. Here we take SEM, banner images, EDM and soft articles as examples.

First, analyze the role of copywriting in

1. SEM: Attract users to click on the landing page.

2. Banner image: attract users to click on the landing page

3. EDM: Title: attract users to open the email, content: make users interested and convert.

4. Soft article: Title: attract users to open the article, content: make users interested and convert.

The copywriting here can be roughly divided into two categories: 1. Attract users to click; 2. Make users interested and convert.

For small companies, copywriting rarely carries the product image, because generally high-sounding copywriting does not convert very well, or it takes a long time to convert. It only deepens the brand's position in the minds of users. Obviously, it is not cost-effective for companies where survival is the top priority.

So, our copywriter decided the direction:

Title: exaggerated and eye-catching.

Content: Write highlights to improve user conversion.

Considering that our product is the first in China, and there are only a few people in the world making it, and it is not well-known in China, so our content tends to focus on usage.

The copy on the banner image has relatively few words, and can only use words like "AI programming increases work efficiency by 3 times."

So the title of our soft article appeared: Artificial intelligence helps you write Java code, 99% of programmers don’t know the necessary code artifact, AI coding artifact makes programmers boil... The title of the email is roughly: Invite you to XXX

SEM titles and descriptions have their own rules. When writing titles, they have their own rules. First of all, they must comply with the rules of common words, which are generally: Are you still looking for a {code completion} tool? Everyone is using AI to write code.

It looks like a clickbait title, but it’s okay to get clicks like this. Of course, if the content doesn’t keep up with the title, people will accuse you of being a clickbait title. The content should still be consistent with the title.

Our content direction is product introduction, the difference and advantages between it and the code completion that comes with the ide. Highlight the convenience and advancement of the product and attach photos of user reviews.

EDM is still narrated in an invitational tone, inviting programming experts to work on artificial intelligence programming together.

5.2. Copy modification

Is everything fine once the copy is written and released? NO! There is still a long way to go, this is just the beginning.

After the copy is implemented, record a series of user data including exposure, clicks, conversions, etc., and analyze problems in the data.

Ideally, we can embed data at every level, create data funnels, and analyze data conversion issues. However, in actual work, very few companies can complete these steps.

Most of the time, it still depends on the keen judgment of operations. As for how to develop judgment, the only way is to step into more pitfalls. But in the end, data that can be counted should still be counted.

Baidu can count the click-through rate, PV entering the website, and the final download volume; banner images cannot count exposure, but can only count the PV entering the website and the final download volume; EDM can count the number of sends, open rate, and clicks on download links; soft articles can count the number of readings, but as for conversion statistics, it is very embarrassing, basically it is sent on the same day, and the registration volume within 24 hours is counted. As for the reading volume compared with the average reading volume of the previous 10 articles, it depends on the title of the article.

After getting these data, decide whether to optimize the title or the copy. There are some fearless operations or copywriting personnel who rely on their feelings to modify anything. Here are some suggestions. If feelings are unreliable, data is more reliable.

The overall copy, title and content were consistent, and after post-optimization, the overall conversion effect exceeded our expectations. With such a long conversion chain, the worst reading conversion effect reached 10%.

6. Activities

After getting the first wave of users, can we only let them be product users? If they are only product users, the cost of attracting them would be a waste. It would be better to turn them into product ambassadors or brand promoters.

Many people will analyze the AARRR model. There is nothing wrong with applying the theory, but the key is to put it into practice.

I personally think there are only two factors that make users share a product: 1. The product is surprising enough to satisfy users after using it, so they automatically initiate sharing. 2. Material rewards to stimulate users to share.

6.1. Purpose of the activity

The pre-selected purpose of this activity: 1. Increase the number of user registrations; 2. Increase product awareness.

We can state the purpose of many activities, but don’t forget that the North Star indicator at this stage is to increase user registrations. All activities and other operations must be aligned with this goal.

Determine the purpose of the activity: increase user registrations!

This purpose has been established, but it cannot be used as an indicator for assessment, so the vague data has to be quantified.

Activity goal : Increase 2,000 users through activities in 1 month. (ps: Goals cannot be set arbitrarily. They must be estimated based on the resources and energy invested, especially for operations with huge KPIs. Be careful that your bonus will disappear if you set goals arbitrarily.)

6.2 Activity Details

If you want users to share the product, it is obvious that the product itself does not make users feel screaming, and the only way to get users to spread the word is through event rewards.

There are many kinds of prizes here, but generally companies will not stimulate users through RMB. Of course, there are also products that stimulate users through RMB. For a tool-type product without a recharge interface, stimulating users through RMB is a bit low.

Fortunately, community version, professional version and enterprise privatization version were planned in the early stage of product design. The community version is free, the VIP version is 128 yuan/year, and enterprise privatization is not within the scope of consideration for the activity.

So we offer 1-year VIP membership to those who invite 3 or more people at the lowest cost, and the membership can be accumulated, and we also give out corporate culture shirts. The publicity channels include free public accounts, soft-article placement and plug-in client prompts.

6.3 Activity Process

Activity process: When users register and log in to the client, they will be given a reminder to upgrade to VIP for free! After the user clicks, he will be redirected to the official website. There is a strong prompt on the official website, which can be seen when the user downloads the client or browses the website, prompting the user to upgrade to the professional version.

By the way, the upgrade to the professional version was not so noticeable at first, but since the conversion rate was so poor, I made it more noticeable and the effect improved.

After the user logs in to the website, there are three ways for the user to choose: 1. Picture invitation, 2. Text invitation, 3. Scan code to share invitation.

After obtaining the invitation code, the user will automatically spread it, and as long as the conditions are met, he or she will be automatically upgraded to a VIP user.

Once the activity goals are set,

Date of activity establishment: July 1, 2019

Activity distribution channels: own official account, soft text channels, official website.

Material preparation: the official account adds a tab for obtaining the professional version; the soft article adds an activity section; the official website adds an activity entrance and prepares materials for sharing; the VIP function is improved and users are automatically upgraded.

When finalizing an event plan, which involves development and material preparation, responsibilities must be assigned to specific people and a last day must be determined. Otherwise, the event cannot be held as scheduled, and operations will definitely be the first to be sacrificed.

6.3. Activity Effect

Monitor the activity results in real time. After the activity went online, only 120 people participated in the first week, and only more than 200 people were invited. At this pace, our goal is far from being achieved.

Of course, a lot of this data cannot be monitored, we can only analyze the entire activity chain to see at which step the user is blocked.

First: The entire link is too long.

It is true that the entire link is long, but users can only be invited after they have registered. Moreover, the current R&D energy is insufficient and there is not enough time to optimize the link.

Second: The prompts in various places are not prominent enough.

Highlight the event in the soft article, create a special welfare section, and mark the key points and the process of obtaining it; the official website highlights how to obtain the professional version; a column in the official account has a special article on how to obtain VIP.

After optimizing these details, within 3 weeks, more than 1,300 users participated in the activity and upgraded to the professional version, and more than 5,000 VIP users were invited.

The entire event process is basically like this: set the event goals, determine the stimulating products, prepare and optimize the product process, determine the distribution channels, determine the event date, and determine the materials to be prepared.

Let me talk about something off topic about the event. When we first cooperated with big Vs to publish soft articles, one of the big Vs came to me angrily and said that we had done group fission. It turned out that he had his own community, and many users who participated in the event sent the invitation codes to his group. I also explained it to him. In the end, he didn't understand, and we both deleted each other as friends. This also indirectly shows that the effect of our activity is still good.

7. Optimization

As channels and activities deepen and the number of users increases, more and more problems are exposed in the products. By the way, we also try our best in product testing. The company’s computers have high configurations, and using them in actual projects and testing them according to test cases are two different things.

We received a lot of complaints from users.

7.1. Collection issues

How to collect user questions? User surveys are one way, but user surveys do not necessarily obtain real data.

We focus on collecting complaints from our own user groups and comments and messages after each big V launch, and collect these issues.

For users who complain in the group, try to have a private chat with them to find out what the problem is. Summarize all the problems and finally select the main ones to solve.

7.2 Product Improvement

There will always be many problems when a product is first launched. As for improving the product, it is especially important in a startup company. After all, there are only a few people, and if they can only do one thing, they cannot do another.

Finally, we defined the two most important issues, latency and lag.

Some of the lag is caused by delay, so solving the delay becomes the primary issue.

Here I would like to introduce a four-quadrant bug repair to quickly define which bugs should be fixed first.

With this bug repair four quadrant, we can quickly define the problem of repair delay. Other problems that cause IDE lag on individual computers are fatal and affect a small number of people, so we will not deal with them for the time being. After all, few people are using low-spec computers now.

Fixing bugs is part of solving the retention problem in the AARRR model. Of course, there are other ways to achieve retention, which we will not discuss for now.

At this point, we have introduced how to acquire seed users from 0 to 1 in product operations. Next is the theoretical part, and we can finally talk about theory openly.

As for whether there is a way to acquire seed users for free, of course there is, but I was not successful in doing so. I only acquired several hundred seed users in one month, and that was before the product was launched. When I have time, I will share with you how to acquire seed users with a budget of zero yuan.

Author: Zhang Mu, Operations Officer

Source: Operations Officer Zhang Mu ( coozhangmu )

Related reading:

How to continuously obtain seed users?

Seed User Operation Methodology

How to find seed users in product operation?

Case analysis: Acquisition and maintenance of seed users!