Operation Growth Practice: 5 Minimalist Cases to Achieve Business Goals

In 2014, I bought a MAC notebook for my notebook. My mom wasn't too happy. The old man is used to thrifting and thought that the Lenovo I was using was solid and cheap, so I used data to speak for myself and made a positive correlation. I said: Old lady, look, I use a Lenovo laptop, and the speed is slow. I basically have to go to bed at 2 a.m. every day. Since I got a MAC laptop, the speed is super fast, and I go to bed at 9 o'clock. The old lady was happy and said directly: Why don't you buy two? Then you can rest after eating.

Parents in the world are all cooks. Who doesn’t care about the health of their children? Just like that, I took the opportunity to buy an iPad again.

This is the persuasive role of data in real life, evidence presented in court.

The same goes for operations positions, which rely on products, technology, news engines, markets, etc. to support them. If there is no data to speak for themselves, it will be difficult to leverage them.

But for those of us who work in operations, data itself is naturally weak. Like me, I didn’t study mathematics well when I was in school, and I don’t understand statistics. I can’t even do the math when I go to the vegetable market to buy vegetables. I have a strong sense of rationality. what to do? There is no solution, I can only practice it every day, so since 2010, I have tried to rely on data to speak and practice my logic skills. I am usually very lazy, but I am very diligent about data. I can read it and memorize it when I have time.

Over the years, I have dealt with data product managers, data analysts, and statisticians. Every time I hear them talk about a new concept, I have to look it up on Baidu, but I still don’t know what it is. Confidence interval, what is probability distribution, I am stupid, so I convinced myself and proposed a concept of minimalist data and optimized source table, trying to make decisions based on an Excel table and simple data. Even data novices can get started quickly. Boost your own business growth. This article summarizes some of the cases in which I relied on minimalist data to improve business target growth over the years.

Let’s first talk about the four major characteristics of minimalist data:

1. Data is extremely easy to obtain, such as WeChat’s backend data, GA data, standard report data made by the company’s data department, etc. wait.

2. The analysis is extremely simple, and you can get insights by just using Excel. I originally liked using GA because it has powerful functions and is free. You can view data in various dimensions and it looks high-end and classy. But since GA was banned and the company no longer used it, I began to think about making a set of local data myself and sorting out all operational projects into a daily table, which I named Optimization Source Table. In this way, I can use Excel's powerful data analysis functions, such as pivot tables, to guide my operations.

3. The idea of ????data operation is extremely simple. Operation is the process of discovering problems and then solving them. Advanced data models are generally used for business decision-making, which are explicit in operations and basically not used.

4. There are certain errors in minimalist data, and there will be personal subjective experience in it. I usually make decisions as long as I have confidence in the data. There is no precise data in the world, only relatively accurate data. I once read a sentence: Making a mistake that can bring you something is far better than doing nothing at all, and it is very suitable for operations.

In addition, in all the cases in this article, the data part has been simplified, bypassing the hidden points, monitoring strategy formulation, implementation, data cleaning and other links. Only the core methods are discussed so that everyone can understand at a glance. Mentioned two core words that I often talk about, trial and error and optimization. Trial and error is the method, optimization is the soul.

Okay, let’s look at the sharing outline of this article:

I often optimize the source table and record the daily data. Sometimes it will be easier if the data department has a standard report. Some.

Optimizing the source table will not bring you much value in the short term, but when the data accumulates to a certain level, you will get a lot of insights from this table. For example, you can reasonably deduce and decompose monthly goals, weekly goals, and daily goals based on the daily performance table of business goals, so that you can truly know your daily operational actions. Because the daily target performance table shows your business rules, especially when you are responsible for the operation of the entire product project and set KPIs for your subordinates, your subordinates will never say that you are patting your head.

Let me give you an example. This is what I learned from the book "Data Management". Since I read this book in 2014, I have applied this technique to Internet product operations.

Following the principle of minimalism, I will only talk about how to decompose it into months, because the ideas of decomposing it into weeks and days are similar. I simply provide an idea.

First, you need to find historical data. I found the daily UV data of PC users for the whole year from the company's data platform. I’ll use UV as an example here, but you can also use LV, Visit, and PV.

Calculate the values ??of the Nth week and day of the week fields from date data. Exclude abnormal dates, mainly holidays, because the traffic of the general website is not normal during holidays. Also, exclude the dates of special promotion periods. This depends on the usual operation diary and keeping a record of daily operation behavior. . Then we get this picture:

We start to decompose this table in perspective.

During the perspective, holiday information and special promotion days are identified, filtered out during the perspective, and not included in the total list.

Did you find it? If you look at the line chart, it's obvious that there are low and peak seasons for business. Therefore, when we complete the annual goals, we must also reasonably distinguish how much is accomplished in the off-season and how much is accomplished in the peak season. Rather than how much is completed per month on average.

Weight is an artificial added value for us to calculate the user. Generally, the lowest value is first assigned a value of 1, such as May in the picture, and then the average UV value of other months is divided by the UV value of May to get the weight of each month.

For example, the total UV in 2015 was 8.8ww, and your boss doubled it for you. Then you can use the weights to equalize how much you need to complete each month. For example, in January in the picture, I calculated the ratio of January's weight to the total weight, and then multiplied the KPI by this ratio to get the KPI to be completed in January.

So, what is the purpose of decomposing KPIs?

1. Make an annual operation plan as early as possible. If the target cannot be achieved in the first month, provide timely feedback to the boss and apply for resources early;

2. Manage the boss’s expectations and control the completion of KPIs Don’t fail to control the rhythm of your work in the first month. If you finish it early, your boss will set a higher KPI for you. Sometimes you dig your own operational pitfalls.

This is an event planned by our collective team, ***5 people, I am mainly responsible for copywriting and data statistics.

At that time, the boss was bold and told the new product team members that you can randomly select 200,000 users as seed users among the existing 40 million users, and then introduce 100,000 initial users to the new product. The user cost should not exceed 5 yuan, just put this sentence down and leave.

We have resources and cost limits. Let’s put aside the cost first. The screening of 200,000 seed users is crucial. This is the ability to segment users. After much thought, we realized that the autumn recruitment was coming soon, and fresh graduates had a strong desire to find jobs. So we carefully selected 200,000 highly active fresh graduate users as seeds to start a prairie fire.

We have planned an activity to promote sharing and attract new users for this group of users. It just so happened that the autumn recruitment for fresh graduates was advanced that year, and many Fortune 500 customers entered the campus recruitment in advance. So our event gimmick starts from these big customers. The theme is that some well-known companies recruit students in advance. If you invite students to join, you can get red envelopes and give you a variety of rewards. There is no upper limit on the red envelopes.

Next, we planned a spreadable activity based on this theme. How did we do it? I sorted out the process during the review. Look at this Excel sheet.

This table is the data of the entire optimization page and effect. I will pick some key points to talk about.

For example, channel strategy, which channel to attract new users from, which part of users will be used for testing, which part will be officially promoted on a large scale, etc. Due to the particularity of the WeChat platform, SMS, app push, page advertising or interception, EDM and other channels are not very user-friendly. We also want to use our own WeChat account, but our WeChat fans have users of all identities, not just fresh graduates. After comparing and comparing, based on experience, I chose a certain channel.

Also, when planning the first draft, the activity process and key node data must be planned and monitored in advance. This point needs to be mentioned. Since it is an event planning, it must be considered carefully. Otherwise, after you go online, the technology may not be able to see the log data, such as the channel conversion rate, the number of shares brought through the channel, and the number of successful registrations. At this time Even if you are busy in vain, I made a huge mistake in this activity.

Don’t promote the event in full volume as soon as it goes online. It is necessary to continuously optimize key pages or key process nodes. After optimizing the conversion of pages and processes to a certain extent, if you feel that it cannot be optimized any further, you may need to use node conversion data to back-end. Based on the obtained conversion rate, we can calculate that we can achieve the goal. Now let’s promote it in full volume.

Because this case is mainly about the conversion rate of the page, I fixed the ratio of channel conversion, sharing and invitation during the first test. For example, the channel conversion rate of this event reached about 32%, and the ratio of sharing to invitations was 1:5, that is, one share can bring 5 registered users.

There are also some conversion data that we have less control over, such as the conversion of the registration process. The registration process is a standardized functional module, and the optimization changes are relatively large, so this optimization is omitted and we do not focus on it. Of course, in our event, the registration process actually cost a lot of users.

After fixing some secondary variables, we placed the optimization goals on two key pages. One page is the page that introduces seed users, which I call the promotion sharing page. There is also a page that when a user shares it with a friend or circle of friends, the friend clicks to enter the page. I call it the registration page. We continued to test these two pages, testing three versions in total until the sharing rate and registration success rate were optimized to the best level. Just go to large-scale promotion.

This link is the most critical and the most tangled. We struggled for a full month. The end result is that through these three large optimization rounds, we successfully achieved our goals. In one month, we optimized the 3rd version of the activity. We think it is worth it, because such activity process and model have been verified by us to be feasible, can be done for a long time, and are particularly scalable. For example, we can build an event backend to replicate our event patterns specifically for different user groups. Therefore, being slow in the early stage is to be fast in the later stage. With the background tools, we can basically do a similar activity every week. This is the value of optimization. Optimization is the cultivation of internal strength, with the purpose of enhancing core operational capabilities.

Now let’s talk about page optimization ideas and how we think about it. In fact, in summary, there are only 3 key points:

For example, in the first version of the sharing page, we put Bole The award turned into a red envelope, and then the sharing conversion increased by 3 points. The transfer to the registration page was even more obvious. We added the invitation element to the button. We no longer talked about signing up as soon as possible and getting the offer quickly. I used the sharer invitation to write an article. This is like your colleague saying to you, "Let's go have a big dinner tonight." You may still hesitate because you are not sure whether he invited AA or whether he suddenly said after finishing the meal, "I forgot my wallet." . And if your colleague says, let’s go, I’ll invite you to a big dinner. If it were me, I would definitely go. So we changed an idea, and the conversion rate to the registration page increased by 10 points.

In our first and second editions, we ignored this element and made a list of famous companies. After thinking about it later, we felt that users might think that this list was clickable, which was misleading. Then users find that they can't click, they feel a little emotional, and their motivation to share and register is insufficient. So in the third version, we eliminated this interference from the visual design.

The key point is in the third version. We changed our thinking. In the first and second version, we only received red envelopes after users invited friends to join and successfully registered. If we allow seed users to get red envelopes by reposting, and force reposting, imagine that the user will feel that this activity is very credible, and the sharing rate will also increase. Maybe he will not only share to the circle of friends, but also go to various groups Sharing is not necessarily possible. Therefore, when the user clicks the Join Now button, a pop-up layer pops up to remind the user that they can get red envelopes after sharing, and they can also get red envelopes if they introduce classmates to join.

Following this idea, we successfully completed the goal and obtained an activity mode. During the optimization process, focus on data, record the data of each activity, and monitor conversions in real time so that every trial and error can be displayed with data. You can see in the picture I just shared, I listed all the key data.

Finally, when we fully launched, the number of registered users introduced was more than 57,000. Why did the number of registered users eventually reach nearly 100,000. Moreover, the final introduction of registered users only promoted the data on that day, and it continued to grow the next day. It’s because the registered users introduced became seed users, and they also started to snowball and repost.

When our APP was first launched, except for passively receiving user feedback through some statistical tools, we were unable to directly reach users. For example, we had notifications of major events, pulled back silent users, conducted surveys, etc. So our product has a PUSH notification product, which I temporarily call a job search assistant. It is similar to the official account built into the app. I will push some information every now and then.

Because our copywriting skills have been tempered for a long time and we already have good experience. I thought that the power of copywriting alone might not be able to increase the click-through rate of push. Is there any other gap that I haven’t found to improve performance? After thinking about it, it suddenly occurred to me that if I knew the user’s preference for push content, , and then push relevant content, can we further increase the click-through rate?

My needs come out: I am eager to get the user's preference for content.

So, how do I gain insight? I needed to work on a trial and error strategy, one of my favorite working methods in operations.

Trial and error is the most reliable means of operation and is the core of the existence of the operation position. The biggest fear of trial and error is unreasonable, so the trial and error strategy is very important.

My trial and error strategy, please look at this table:

A brief explanation is as follows:

Trial and error users’ preference for content at different points in time. I set my goal on summarizing monthly patterns. If your business has a pattern, you can put it on a quarterly or even weekly basis.

Copy click-through rate. If users have regular preferences for content, they will definitely have a relatively high click-through rate at a certain point in time.

Finalizing the trial and error variables is a key node for the rationality of the trial and error strategy. The fields marked in orange in the above table are variables.

Sort them in advance. My trial and error was with content types, not individual articles. What are content types? For example, portal channels include technology, entertainment, military, news, etc. I concentrated on selecting 7 categories of content. For example, interview strategies, online application strategies, resume strategies, etc. should be planned in advance and reasonably arranged every week. Step-by-step releases. For example, I post information summary posts every Monday, resume guides on Tuesdays, interviews on Wednesdays, and so on.

The time point is to post at a fixed time every day. For example, I always choose to post at 8 o'clock in the evening. As for how this time point came about, I got it through research. I have a WeChat account for testing copywriting. Because I usually answer questions from fresh graduates, they are very nice to me. I sent a survey post asking them when is the appropriate time to push messages to them, and 80% of them said 7-9pm. point.

Friends who have worked on APP notification products know that iOS does not count the received data. In order to be more scientific, we only use the Android client for testing. The copywriting click-through rate depends on the click-to-receive ratio, which is more reliable than the click-to-distribute ratio.

Find someone to try and make mistakes. This is very important. I now know that most students are not too bothered by users who are fresh graduates as long as they push relevant job search information. Because job hunting is a necessity, if the job search is successful, most of them will turn off notifications or uninstall the app. But I didn’t know that at the time, and I felt that pushing information was very disturbing. Therefore, in order to prevent users from complaining by sending messages every day, I only select users with certain characteristics, such as highly active users, that is, those who have logged in and made deliveries that day; or silent users, which can reduce harassment even more , because if the silent user is activated by my push, he will immediately return to the active pool and will not receive my push. Moreover, the reason why the silent user is silent may be because he has no demand. If he clicks on my push, it proves that he likes my content. It just suits my purpose of trial and error.

As we all know, the title of Android can be customized, unlike iOS, the title can only be the brand name. Because my purpose is trial and error, not to get the click-through rate effect, I don’t customize the title and just write the brand name directly. It's more controllable this way.

The most difficult variable to control. The title party has the greatest click effect, but this is an abnormal data of trial and error. I must ensure that the level of copywriting is within a certain intensity range in order to reasonably obtain business understanding. This again involves a trial-and-error strategy of copywriting intensity. This is more complicated and will not be discussed today.

In my original work experience on WeChat and content channels, I have tried and made a summary of the strength of copywriting. I know which copywriting is high-strength and which is low-strength, because I am the one doing it myself: I record data every day. So I can reasonably know the normal click level of each copy. Everyone just needs to remember that there are levels of copywriting intensity. This case uses Level 2 copywriting intensity, and under normal circumstances the click-through rate fluctuates by 2 points.

Of course, if you don’t have a copywriting level, you can also compile a few articles for each content type, and then work on segmenting the target users, such as taking the users who logged in on the day and setting up a filtering mechanism: each user Only receive once during the trial and error period, etc. In this way, you can push these articles to the same type of users.

Pay attention when cleaning data. For example, sometimes the sending channel is unstable, not sent, or the statistics are wrong. These abnormal data must be eliminated during cleaning.

In this way, I finalized 7 variables that affect my business perception. Strive to get reasonable and scientific advice!

One year later, the trial and error was over, and I began to summarize the rules. I saw through this table:

Do you see it? The types of content highlighted in pink are definitely the most popular content that month. So, in a certain month of the next year, if I increase the content that users like most, can I improve my performance?

This is my content trial and error strategy. The conclusion is simple, but the process is complicated.

Of course, my trial and error case is a long-term battle, because we are a mature product. If your product is in the start-up or growth stage, you can choose short-term trial and error, as long as you control two points:

First, be sure to clarify the trial and error goals, and the goals must be unique. Because trial and error is to gain business understanding, not trial and error for the sake of trial and error;

2. During trial and error, you must find ways to determine the variable factors that affect your trial and error conclusions, and strive to minimize the impact on the business Cognition.

This is a case from 2015. It was mentioned in my article "Murong Xuefei and I talked about operations late at night". The writing was relatively rough. Today I sorted out my thoughts again and made it organized. Let me tell you about it and release my optimized source table for WeChat operations. In addition, everyone should pay attention to one thing: the number of shared collections counted by WeChat in 2015 was put together. Unlike now, the statistics of sharing and collection are separated, making this method of locating content more scientific.

When I first took over a WeChat ID content operation, the daily net increase in fans was negative. So how did I solve this problem? I still work hard on the content. The leader is very anxious to see results, and I don’t have time to do content trial and error strategies. Then I can discover patterns from historical data. Let me briefly describe it:

To clarify my business problem: WeChat’s daily net increase in fans is negative.

Analyzing the problem, there are actually two solutions to this problem:

1. Saving money, this is a content selection problem: what kind of content should I make to meet user needs? Keep users from running away. That is, making a fuss about existing fans to reduce the loss of fans;

2. Attracting new fans. If fans continue to lose, but I increase the number of daily fans more than the number of fans lost, then the daily net increase of fans will be Isn’t it right? The number of daily fans and the number of retweets are closely related. Generally speaking, there is a positive correlation. The more retweets, the more followers you gain. And forwarding is closely related to the content.

Therefore, my business problem is actually content positioning. I need to find users' preferences for WeChat content. My idea is very simple.

As shown below, I made a reading order for this picture.

I have omitted the number of senders because it involves business privacy

This table is the true appearance of one of my optimized source tables. I have a habit. Every time I work on an operation project, whether it is the operation of the entire product or as small as a navigation category, focus map, recommendation position, etc., I will figure out a table like this myself to find problems and summarize the rules. , and then guide operational behavior.

It took me two days to manually collect all the historical push data of the past year into this table. Even more tiring is categorizing them one by one. I roughly divided these contents into 10 categories. Then classify each article into these 10 categories (classification fields in the table), and then this source data table comes out.

1. The number of new followers (fuzzy) field is the number of fans added the next day. It is not precise data and is for reference only. If your WeChat ID only posts one piece of content every day, this data will be more accurate.

2. If there is no fixed number of images and texts in your historical push, you need to roughly calculate the ratio of retweets and followers, and then attribute the increase of followers to a single article based on the number of reposts of a single article. In pictures and texts. This is an arduous project that requires you to clarify the data piece by piece and exclude outliers that are difficult to control.

3. If you don’t want to follow the second step to avoid the trouble of cleaning data, you can also formulate a short-term content trial and error strategy according to the previous part, and you can send a single graphic message every day. This trial and error variable has solidified.

Then, through perspective analysis, my WeChat ID content operation strategy came out, as shown below:

If you look at the "Determine content type" column, you will find that salary, workplace , inspirational, skills and other content exceeded the average number of followers. The problem is solved: I focus on these four categories of content. Then I studied the titles with high reading numbers, and tried to increase the number of readings for each article. In about 10 working days, I successfully turned the number of daily net followers from negative to positive. ”

Many people like me, who are engaged in WeChat content operations, see other people’s content with good reading numbers and post it. In fact, this is not the most efficient. Other people’s fans may not have the same characteristics as your fans. Content requirements. The same article can reach 100,00 on other people’s WeChat account, but it may not achieve the same effect on your WeChat account.

After I corrected the daily net increase in followers, What did you do next? After the correction, the account was in a healthy operating state. I was not willing to be just an operations manager, so my next goal was to increase the number of followers. The next issue was the operation strategy. , not shown here!

Finally, in the fifth case, let’s talk about the fundamental question, what is the value of data in real work? Why does operations need data analysis? I summarize it as follows:

1. Find problems: Problems are the driving elements. We must constantly monitor product performance, analyze data, find out problems that affect business goals, and eliminate them

2. Reduce costs: John. Wanamaker said: Half of my advertising money is wasted, but I don’t know which half. Analysis can effectively reduce costs.

3. Decision-making basis: In enterprises, big bosses generally need it. Rely on data to make investment decisions, marketing decisions and strategic decisions.

In a word: analysis can monitor effects, find problems, gain insights, control costs, and achieve the goal of maximizing business value. Is it right or wrong? What I said is correct, but in most cases we need to be grounded. As I mentioned at the beginning, in real work, after many years of experience, the data is the biggest. The function is to "communicate evidence", just like if you commit a crime and refuse to admit it, and then the lawyer collects evidence.

This is the fourth function of data:

4. Communication. Evidence: Analysis can provide evidence to effectively decide on operational plans. Operations must monitor and optimize the business performance of each department, and then provide effective evidence to convince relevant business departments to achieve our goal of optimizing products and services. p>

Sometimes, with the naked eye or by personal experience, you can tell that there is a problem, but the relevant business departments will not change it and let the data speak for themselves. Then we will try our best to find evidence.

For example, when I optimized the conversion rate of major processes on a recruitment website, I experienced it countless times and found it to be really cumbersome. However, if the product does not change, I need evidence. There is also a second optimization case. I did not talk about the registration process. Conversion, in fact, the loss of that process is too complicated.

Then, in order to find evidence, you need to find data, such as the funnel chart below. Figure.

See, there are too many leaks in the 4-step delivery process. The loss ratio of each process node is "unattainable", so I quickly attached the analysis and guidance below. Business secrets will not be released. In fact, it’s clear at first glance that it’s just a matter of wording and written more tactfully. Use descriptive words such as "very, too" with caution, otherwise the product will be angry.

After giving the product, the product was still struggling with the accuracy of the data. Then I thought of another method. The experience data method came in handy. I wanted to see what the major processes of my competitors were like. of? I decided to compare based on the user task load and the time it took to complete the task. What impressed me the most was that it was the summer of 2014. People are particularly irritable when it is hot. I was shirtless and humming alone to experience the three major processes of competing products and record the data. The final results are as follows:

For the three major processes, the time it takes users to complete tasks and the number of actions are significantly higher than those of competitors. This is not a good phenomenon, and also confirms the reason for the high churn rate mentioned above: there are many recruitment platforms. , users waste time and effort in order to submit a resume, and their willingness to transfer platforms will be very high.

Finally, I used the funnel chart tool and experience data to complete an optimization certificate. Submitted product for review.

Again, the data may not be accurate, but a mistake that can make people gain something is far better than doing nothing at all. I would like to encourage everyone.