How much should product and operations professionals know about data analysis? We gave the following answers

Data analysis capabilities are important for both product and operations personnel. How important are they? Let’s go straight to the data.

We used Python to crawl 500 job requirements for event operations , content operations , and user operations from 51job.com, performed a word frequency analysis on them, and produced the following chart.

We have come to the following conclusion: Employers generally believe that data analysis capabilities are very important for operators (of course, the more important core competitiveness of nuclear operations is product thinking and marketing planning capabilities). However, what’s interesting is that many operations staff focus too much on their marketing capabilities (such as copywriting skills and event planning skills) but ignore the improvement of data analysis capabilities. The team I lead also has this problem, so I wrote this article for your reference.

The role of data analysis in operations

Operators are the people closest to the business. Having efficient data analysis capabilities helps us quickly make operational decisions that are highly relevant to business growth. The data analysis done by excellent operators will have more practical guiding significance for the business. It will not be superficial and will not degenerate into simply "collecting data", "making tables" or "writing reports".

For sales and operations in the Internet era, data analysis has three main functions.

1. Describe the current product status and user status in a concrete way, identify problems, and help make operational decisions;
2. Verify whether the operational strategy is effective;
3. Explore and predict future possibilities to optimize products and operations;

These three functions are also progressive, mining data from existing behaviors, inferring behaviors through data, and then predicting the future through data. Data analysis cannot be separated from products. All analyzed data comes from products and user behavior , and the analysis conclusions serve the products and activate user behavior.

Analytical thinking

Growth formula thinking

To change the motion state of an object, there must be the existence of force or field. The growth of product scale and user growth must have its growth engine.

Enterprise growth = coefficient 1 factor 1 + coefficient 2 factor 2 + … + coefficient n * factor n

Finding the factors that drive the business through understanding the business is a matter of experience. It is determined based on our familiarity with the business, user sensitivity, and understanding of marketing. We use rapid iteration and experimentation to verify whether the various factors we have selected are reasonable.

Let's talk about the factors first, and give an extremely simple example:

Revenue - Expenses = Profit

Corporate profits have declined. What is the reason? The core driving force is a decrease in revenue or an increase in expenses.

But don’t forget that there is a coefficient before the factor, because there are so many factors that affect the core business, we should find the key factors . This coefficient describes the degree of influence of the factors on the core business.

Here's another extremely simple example:

The mall’s turnover = the mall’s revenue on the negative first floor + the mall’s revenue on the first floor + the mall’s revenue on the second floor. The negative first floor is the mall’s parking lot, the first floor sells men’s and women’s fashion clothing, and the second floor is a food court. We added the coefficient based on our personal experience, and the mall turnover = 1 mall minus 1st floor income + 30 mall 1st floor income + 5*mall 2nd floor income. The specific reason is that clothing malls have high gross profit margins, and people come here for the mall’s core business. Therefore, the income on the first floor of the shopping mall becomes the most critical factor. When there are too many factors to consider, factors with large coefficients become the key factors that we need to consider first. What we are talking about here is not a mathematical formula. The plus sign in the growth formula refers to the organic superposition of growth factors, rather than a simple mathematical addition.

Pyramid Thinking

The pyramid principle has a core rule: mutual independence and complete exhaustion. It is an excellent way of thinking and expression. Independence of each other means that each sub-argument should not conflict or couple with each other and should all be independent modules. Complete exhaustion means that all sub-arguments are presented without any omissions. In the early stages, it is difficult for us to be completely exhaustive, but we must think with this mindset.

One day, my subordinate came to me to report and said:

Brother Hao, there are only more than 30,000 users participating in this event, and the registration conversion rate is only 30%. Recently, product conversion has not been good, the server often crashes, and the channel guidance for registration is weak. It seems that user demand has also decreased, and the actions of competitors have also caused some users to run to them.

After listening to it, I had a very cute look on my face. What are you talking about, kid?

It is difficult for our brains to remember multiple independent arguments at the same time. If we connect them together with a certain logic, the people listening to you will understand your point of view.

According to the idea of ​​mutual independence and complete exhaustion, we can first list the points he reported:

• 30,000 people participated in the event (is this a lot or a little? We need to compare it with past data for analysis)
• Registration conversion rate 30% (Is this a lot or a little? Need to compare with past data for analysis)
• Poor product conversion
• Server downtime
• Weak channels
• Declining user demand
• Competitor Behavior

We are organizing it using the pyramid method. In fact, the core idea of ​​the reporter should be that the sales of the products have declined recently. The others are just superficial phenomena that support the conclusion of the decline in sales. There are some possible reasons. We operate the two growth formulas of sales = new customer sales + old customer sales and new customer sales = new customer traffic + new customer conversion + new customer unit price. We find out the key factors of growth as traffic, traffic conversion and old customer repurchase and organize the following pyramid.

Categorical Thinking

User segmentation, market segmentation, product segmentation, we use classification thinking everywhere when making operational decisions. There are commonalities and differences between things. The basic idea of ​​classification thinking is that we can separate things with very different core indicators. As mentioned above, for the business growth factors, we can classify the relevant key factors.

Through the two mutually restrictive factors of sales growth rate and market share, the Boston Matrix classifies corporate products into star products, cash cow products, problem products, and dog products, and then analyzes and plans the corporate product portfolio to achieve the company's profit goals.

Funnel Thinking

The funnel model is a panacea for product operation analysis. From user entry to final conversion, there will be loss in each link, each link will have a conversion rate, and the number of people in each link decreases successively. Each path of the user forms a funnel.

There are two key points to funnel thinking. First, we must pay attention to the loss at each step of the funnel, analyze the reasons behind the loss at each step, and gradually reduce user loss . Second, we need to consider not only the reasons for loss, but also the relationship between the upper and lower levels. For example, in order to attract new users , a product uses misleading copywriting such as "Sign up and get an iPhone" to entice users to come in. Although this can bring in a lot of traffic in the first stage, if users find that the product is not what they advertised, it is likely to lead to a very low subsequent conversion rate, and make users feel bad, leading to negative comments about the product.

Should understand analytical tools

Always remember that we are operations or products, not data analysts. With limited energy, you need to be proficient in two tools, one is Excel and the other is PPT. Excel is mainly used for data processing, data cleaning, and data visualization, while PPT is mainly used to display data analysis results and write reports to guide operations.

For products and operations, the ultimate goal of data analysis is to solve problems. Don't just pursue good-looking charts and advanced data analysis methods. Mastering 20% ​​of data analysis methods and tools will be able to solve 80% of data analysis problems.

Data analysis process

We can define data analysis as: using appropriate statistical methods to summarize and develop large amounts of collected data in order to extract information, form conclusions, and guide work.

I think data analysis should have the following process:

1. Clarify the purpose and ideas: What problem is this data analysis intended to solve?

This is the first step in data analysis. We must bring questions to find answers . The amount of data is huge, and the data are interrelated. If we don’t bring questions with us, we will get lost in the ocean of data.

Not only do we need to have questions, we also need to have the right questions when we go out. Here is an example.

• Bad question: Why hasn’t the number of orders from new users increased? How can we improve the conversion of new customers?
• Reasonable question: Recently, we have taken offline the function of automatically sending new user gift packs after users register. Has this led to a decline in the conversion rate of new customers?

After clarifying the purpose, you need to determine your own analysis ideas. The analysis ideas mainly include various business analysis models and marketing analysis models. These business models are the core competitiveness of our operations. Compared with data analysts, we understand marketing and products better, which will not be elaborated here.

"Who Says Newbies Can't Analyze Data" mentions some commonly used marketing management methodologies.

• PEST analysis method: used to analyze the macro environment, including political, economic, social and technological aspects.
• 5W2H analysis method: Why, What, Who, When, Where, How, and How much .
• 4P marketing theory: analyzes the company's overall operations, including four major factors: product, price , channel, and promotion.
• User behavior theory: mainly used for website traffic analysis, such as returning visitors, new visitors, churn rate, etc., select some applicable indicators from many.

2. Collect receipts: Find data related to the problem from the site database or external sources

Human behavior every day generates massive amounts of data. When you open your eyes, your weight, height, heart rate, and blood pressure are all data. The temperature, humidity, and PM2.5 outside are also data.

So where do we find the data we need? From macro to micro, we divide the data sources into the following five stages: macro data, corresponding industry user data, Internet user data, similar product data, and proprietary product data. Among them, comrades in product and operation need to pay special attention to the corresponding Internet industry data, similar product data, and their own product data.

3. Data processing and cleaning

Data cleaning refers to the final step of discovering and correcting identifiable errors in data files, including checking data consistency, handling invalid values ​​and missing values, etc.

Here are a few examples to illustrate. The first is data consistency: based on the reasonable value range and mutual relationship of each variable, check whether the data meets the requirements, and find data that exceeds the normal range, is logically unreasonable or contradictory. For example, a male may have a gynecological treatment record. For this type of data, we can take out the data source and re-verify it, and sometimes we need to delete it directly. Invalid values: The user's height is a negative number, and two pieces of data are completely duplicated. These can all be considered invalid values. Missing values ​​are just as the name implies, missing values. For invalid values ​​or missing values, we can estimate or delete them.

Cleaning data using deduplication

4. Build data model and analyze data

Finally, the real data analysis began. Yes, I am not kidding you. Data analysts spend more than 80% of their time every day collecting and cleaning data that is suitable for analysis. The data analysis process is mainly like this.

• Observe the data and see what the current product status is?
• Why is this happening? What changes have taken place in the overall environment? What actions did we take?
• Determine what might happen next?

Data analysis has some basic analysis methods. If we are proficient in using these data analysis methods, we will be able to answer the above questions by studying the data.

Comparative analysis

Compare two or more data and analyze their differences to reveal the development laws of things represented by these data. We often hear about horizontal comparison and vertical comparison. The comparison of different indicators under the same time conditions is horizontal comparison, such as comparing the GDP of China, the United States, Russia and Japan. Vertical comparison is to compare the values ​​of different periods under the same conditions, such as the comparison of my country's annual GDP.

When analyzing data, it is particularly important to choose an appropriate comparison system.

• Comparison with target, comparison at different times (month-on-month, year-on-year)
• Comparison of different entities (e.g., comparison of conversion rates of different traffic channels)
• Industry comparison (comparison of traffic conversion rates of different products in the same channel)
• Comparison before and after the operation (comparison between users who received coupons and users who did not receive coupons)
• Compare with the average or median (elementary school students like to compare their grades with the class average)

Through comparison, we can judge the situation reflected behind the indicators and determine the current status of the product.

Growth formula and weighted analysis method

As we mentioned earlier, core indicators have corresponding growth formulas, and the weight of each corresponding growth driver is different. Here we introduce a simple method for determining weights - the target matrix method. The working principle of the target optimization matrix is ​​to simplify the fuzzy thinking of the human brain into the 0/1 thinking of the computer and finally obtain a quantitative result.

The goal matrix mainly puts the decision-making factors in a matrix and allows more experienced colleagues in the team to determine the importance of each factor.

Next, let’s take an example. Suppose your criteria for choosing a spouse include the following factors: owning a house and a car, handsome, highly educated, good character, and being together for a long time. We create the following matrix:

• Comparing having a house and a car with being handsome, having a house and a car is more important. Enter 1
• Comparing owning a house and a car with having good character, owning a house and a car is more important. Enter 1
• Comparing having a house and a car with having good character, having a house and a car is not that important, enter 0

After the comparison of the house and car is completed, compare the other items one by one and fill in the total:

Make corrections to the 0-point item, such as adding 0.5 points to it. And calculate the weights:

Finally, calculate the total/total of all indicators*100%, and the calculated value is the weight value of this item.

Matrix analysis

Matrix correlation analysis is a vivid and easy-to-use analysis method. Matrix analysis correlates two or more important indicators. The matrix analysis method can mainly solve the decision-making problem of how to allocate resources and specifically identify the key areas where the company needs to improve in management.

The matrix analysis method mainly establishes a plane rectangular coordinate system, in which the two coordinate axes correspond to the two attributes of things respectively.

For example, the channels we often use to communicate with users are: SMS, APP push , email EDM, in-site messages, and homepage pop-ups. If we can only choose two channels for connection due to limited development resources, how should we choose? There are two key elements in message communication: cost and information reach. These two parameters are used to establish a coordinate system. The coordinate system shown in the figure below is obtained , and the four quadrants correspond to the following attributes:

Based on our analysis, we placed several channels in the above quadrant table according to their performance.

After comprehensive analysis of the various points in the above figure, we can see that SMS has a far higher information reach rate, but the cost is very high. Therefore, SMS should be used to recover lost customers because they may have uninstalled the APP. Other channels with low reach rates may not be able to reach these users, so we have to use higher costs to reach them. The cost of APP push and in-site messaging is low, but the effect of reaching inactive users is poor, so we can use these two channels to communicate with active users. The home page pop-up window, on the other hand, is a channel with relatively high-quality data and is suitable for promotion to all users.

5. Draw conclusions and make decisions

In our country, decision-making is a special process, generally a collective decision, but the decision-making power is mainly concentrated in the hands of a few top managers. Grassroots managers rarely have the power to make decisions. Once a decision is made, subordinates must strictly implement it. Most of the product or operations friends who are reading this article are middle or lower-level managers, or even just executors.

Therefore, when we draw conclusions, they must be conclusions that our superiors can quickly understand and comprehend. When reporting, the lengthy data analysis process should be summarized into several independent and substantive conclusions.

6. Report Writing

When you have completed the data analysis in the above stages, congratulations, you have reached the last step, report writing. Report writing is the only way to present your data analysis ideas and conclusions.

A bit like the three-paragraph essay in the Chinese language test for the college entrance examination, the report should have the following parts:

Here are a few key points from the analysis report:

• Conclusion first.
• Do not write redundant data, and each icon presented must give a corresponding conclusion.
• Conclusions that can be seen at a glance do not need to be written out. For example, if there is an obvious difference between the two bars in a bar chart, there is no need to add additional text to explain the obvious growth.
• There must be a landing point: There must be corresponding solutions for the product problems or deficiencies presented by the data.

Always remember, we are product and operation, not data analysts. We should focus on conclusions, actions and measures.

Business indicators that operations need to pay attention to and understand

Basic traffic indicators

Traffic indicators are the basic indicators in Internet operations . Traffic includes several indicators. The following are the most basic business indicators:

• PV (page view) is the data generated by accessing the page. If a user visits 5 pages, 5 PVs will be generated.
• UV (user view) is the number of visitors to a specific page. No matter how many times an account clicks on a page, the UV is 1 because there is only one visitor.
• IP: The network IP number for the entire site. You logged into this website using your home computer, and then your cousin also logged into his account using the same computer and visited the same website, but this time the IP address was still only 1, because you and your cousin used the same computer and the network had the same IP address.
• Page dwell time: Dwell time refers to the length of time a user stays on a website or page.
• Bounce rate: Bounce refers to the situation where users leave the website without clicking on the second page after reaching the landing page . The bounce rate refers to the proportion of visits that directly bounce out of the visits that use the landing page as the first entry page. The calculation formula is: Bounce rate = Bounce visits / Landing page visits
• Conversion rate of each process: such as registration conversion rate, product details page conversion rate, shopping cart conversion rate, payment conversion rate, etc.

Business Indicators

• Order quantity, order amount
• Each order amount = order amount / order quantity
• Unit price = total sales amount of goods / sales volume of goods
• Average order value = total sales amount of goods in a time period / number of users who place orders in a time period
• GMV: Platform-based e-commerce businesses will focus on GMV (Gross Merchandise Volume), which is the total transaction amount.

User Operation Focus Indicators

The main routine of user operation is user life cycle analysis, which is the data analysis of the user's entire life cycle from inflow, registration, retention , conversion, activity, and loss.

When users register, the main data that need to be considered are the effectiveness of each drainage channel and the unit price of user registration , as well as the user's bounce rate and page dwell time in each registration process . The main purpose is to analyze the pros and cons of each channel, the smoothness of the registration process, and various possible problems. After registration, you should pay attention to user retention, including retention rate , user return visit frequency, core function usage time , etc.

Users who do not convert are not good users. The number of paying users , the proportion of paying users, the growth rate and the conversion rate from registration to payment are all things we may need to pay attention to. We also need to pay attention to the amount of payment, the frequency of repurchases, the average order value, etc. At the same time, we also need to pay attention to the behavior of users who have been active but have not converted.

Indicators for event operations

For every activity, we can operate it as a new product. Activities are operational means to promote a sudden increase in various product indicators in the short term. To judge whether an activity is successful, we must look at the increase in the target indicators. Taking e-commerce activities as an example, the increase in this target indicator may be the conversion of new users' orders, the average order value of new users, the average order value of old users, etc.

We also need to analyze the cost of each channel, the number of traffic generated by each channel, and the number of conversions in each channel , and finally calculate the ROI of each channel to determine which channel has a better effect on activity traffic and conversion.

Content Operations Focus on Indicators

What content operations need to consider is the traffic that the content can bring and the ability to monetize the traffic.

The content itself can attract a certain amount of traffic, and as users spread the content, the traffic will increase exponentially. Finally, we also need to convert the traffic into cash. I believe that content operations need to focus on the number of clicks on the content, the page dwelling time on the content page, the bounce rate of the content page, and the number of likes . The above four indicators can effectively judge whether the title of an article is attractive, whether the content is valuable to users, and whether the content is clickbait . Valuable content may not necessarily be content that users are happy to spread. We also need to pay attention to the amount of content forwarding.

When we have accumulated enough traffic, we also need to consider the conversion and monetization data of the content. The conversion data of the content varies according to the product form, and can be reflected in the number of clicks on paid links, the number of clicks on page ads, the increase in sales of the promoted products or brands during the promotion period , etc.

Different products will have different indicator systems, which cannot be listed here. The core idea is to focus on the user's conversion path in the product, and expand the data indicators that need to be considered from the core conversion path.

Assist in establishing BI system

The BI system is mainly for operations and products. Not all operations staff have the ability to view the database. Large companies with detailed division of labor will not allow operations staff to obtain database permissions. It is also inappropriate for operations staff to spend too much time searching and cleaning data.

Therefore, we need to establish a data dashboard and data analysis system. The data analysis system is an internal product whose users are mainly products and operations. It is mainly led by data product managers and developed by data development engineers. The main purpose is to allow operations staff to easily and conveniently view the core data they care about most and make operational decisions in a timely manner. The BI system can be developed internally by the company team or using third-party tools such as Sensors, Tableau, etc.

How to improve data analysis capabilities

• Look at more data: Get to the office early every day, look at the data reports, and think about the reasons behind the data fluctuations. Over time, you will become a data master. I once saw on Zhihu that the way for data analysts to improve their data analysis skills is to memorize data. Although it is a bit extreme, it does make sense.
• Familiarity with the business: Data analysis is based on business data analysis. Operations and products must be very familiar with the business in order to find existing problems through the data. This is also where we have an advantage over data analysts in the data analysis process.
• You need to be proficient in Excel and understand other tools. In addition to commonly used Excel functions, you also need to be proficient in various icons and data visualization tools, pivot tables, etc. You also need to understand the database language SQL, and it would be even better if you know Python, so that you can communicate with data analysts more quickly, and you can also search and mine some simple databases yourself.

Final Thoughts

Operations is a management discipline, and the improvement of management capabilities mainly lies in practice. This article can only provide you with some practical ideas and methodologies, and the examples in it are relatively simple. Everyone needs to combine ideas and methodologies with their own Internet products and expand the framework to master data analysis more effectively. A journey of a thousand miles begins with a single step. Dear operators and product people, let us encourage each other.

The author of this article @梁彦豪 is compiled and published by (Qinggua Media). Please indicate the author information and source when reprinting!

Product promotion services: APP promotion services Advertising platform Longyou Century