Teach you to understand the essence of the problem from the data.

Thoughts on Innovative Thinking Structure and Work Efficiency

Students outside

February 28(th), 2020

How to gain insight into the essence of the problem from the data? These three tricks teach you to share dry goods.

Warm reminder: text * * * 4977 words, estimated reading time 13 minutes.

In the public examination (administrative professional ability test) line test, I accidentally saw an interesting question to the effect that:

During the Spanish-American War, many Americans were unwilling to join the army for fear of death, so the US military issued a poster that read, "The death rate of the US navy is lower than that of new york citizens."

According to statistics, the death rate of new york citizens is 0/6 per thousand people1.6%, while even in wartime, the death rate of American navy soldiers is only 9 (0.9%) per thousand people.

If you are the target audience of joining the army, what is your reaction to seeing this data? Possible reactions:

1) It turns out that the war mortality rate is so low that I should join the army.

2) The military must have fabricated data. How can the death rate of soldiers be lower than that of ordinary citizens?

So what is the actual situation? Navy soldiers are strong young people, and most of the dead in new york are old, weak and sick. These two data cannot be compared together.

This is a story that we are "manipulated" by data.

There is another story.

1948 During the Liaoshen Campaign, Lin Biao led the Northeast Field Army to conquer Jinzhou, then went north and joined forces with more than 200,000 Liao Yaoxiang Group in western Liaoning.

Lin Biao has a habit of letting the personnel on duty report the data of each unit every day.

One day, when reporting, Lin Biao found that the ratio of captured short guns to long guns, the ratio of cars to carts, and the ratio of captured and killed officers and men were slightly higher than those in other battles.

As a result, Lin Biao was launched, where the enemy was in command, and Liao Yaoxiang was finally captured alive successfully, which achieved a key victory and greatly shortened the war time.

This is a story about discovering opportunities from data and changing history.

My first job was data analysis, and then I went to a consulting company. Because of social phobia, my ability to deal with customers is very poor. Later, because of a data analysis report, customers and partners were impressed, and I found the confidence to continue consulting.

So I've always wanted to write data thinking systematically. Today's article will make a systematic summary of my experience and methods of data thinking, and how to see the essence of the problem from the data literacy and data table.

Have basic data literacy

1) has basic statistical concepts.

Let's start with the most basic concepts: average, median, percentile, mode, deviation, variance and standard deviation. I won't go into details here, but simply talk about the difference between the mean and the median.

Average value: that is, average value. The advantage is that the average is related to all data, but the disadvantage is that it is easily influenced by extreme values.

For example, you and your three friends and Bill Gates form a team, and then the average value of this team is $20 billion. Do you think you have money?

Median: only related to the middle data. The advantage is that it is not affected by extreme values, and the disadvantage is insufficient sensitivity.

Whether you are Bill Gates or unemployed Xiaoming, the median may be similar.

You can recall and learn other concepts by yourself.

2) Avoid data logic errors.

Common Data Logical Fallacy 1: Relevance as Causality

"Some research results show that people with high face value also have high income."

After hearing this conclusion, do you think you should go for plastic surgery?

But it may be because people with high face value are relatively confident, and confident people are easy to succeed in the workplace, so their income is high. In other words, you should probably not have plastic surgery, but should improve your self-confidence.

It is also possible that people with high incomes have the ability to dress themselves up, so they look more attractive. Therefore, the above statement is only about the relationship between face value and income, and does not say that the two are causal.

Common data logical fallacy 2: Feeling is fact.

Vanke, as everyone knows, you may have eaten Haitian soy sauce. Which do you think is the higher market value of these two companies?

If I didn't ask deliberately, most people would think Vanke is much bigger, because it is more famous, and the house is always more expensive than soy sauce, right?

But in fact, the two are almost the same, both of which are about 300 billion. What if I add Yili shares? It seems that we always drink more milk than soy sauce, but the market value of Haitian flavor industry is still higher, almost doubling (300 billion vs.180 billion).

Common data logical fallacy 3: the individual is a whole

When Pinduo first came out, many people couldn't understand it. Then it advertised on CCTV that it had "300 million users", and many people said, "No way, there are thousands of friends in my circle of friends, and none of them is useful. 300 million is fake. "

But the problem is that our circle of friends are all people similar to us, and this sample does not represent China. Therefore, the data that Wang Xing mentioned later that "only 4% of people in China have a bachelor's degree" shocked many people.

Data communication and expression: how to tell stories with data

I once wrote an article about the importance of telling stories.

But the story can't just have a framework, but also have content and material. If you have enough data literacy and know how to present data and express data at the same time, then you can incorporate enough convincing data into the story, and the story will naturally become very convincing.

1) Understand the purpose and object of communication.

If you persuaded a customer to buy your wealth management products, what would you tell him?

The first type: this wealth management product has a loss probability of 10%.

The second type: this wealth management product has a 90% probability of earning.

The latter, of course. He is willing to buy it, but if it is the former statement, he may be scared.

Therefore, when you communicate with different people in the company, you should also present different data.

For example, the top management may be concerned about the company's overall revenue, profit and other related data, the middle management may be concerned about the KPI data of their own departments, and the supervisor is more concerned about the success or failure of an activity or an initiative.

2) Select the appropriate data expression type.

I often read the data analysis reports given to me by my colleagues and find that most people don't know how to present the data.

For example, if you lead a customer service department, two people left the team some time ago, and the workload increased, so the phone could not be answered at all. You want to apply for two more people from your boss, who asks you to take stock of your monthly workload. How do you present these data?

It is nothing more than comparing the incoming electricity and the timely answering volume, so most people present it like this:

However, a better data expression should be like this:

How to use a more suitable data chart type?

In the past, when I was doing data analysis and consulting, because I often had to make charts, I collected many types of charts, and there were more than 100 visually, but in fact, after the final inventory, there were no more than 10 commonly used for so many years. The scope of application is as follows:

Scatter chart (suitable for correlation), line chart (suitable for trend), horizontal and vertical bar chart (suitable for contrast), waterfall chart (suitable for evolution)

Heat map (suitable for focusing), radar map (suitable for multi-index), word cloud map (suitable for distribution) and so on.

When we are thinking about the core ideas that need to be presented, we need to think about which one is appropriate through these pictures, instead of bringing a pie.

3) conform to the principle of data visualization

Visualization of data is also very important, because without visualization, it is just a string of numbers, which is no different from text information.

Several principles of data visualization: don't read too high a threshold, don't use too many colors, highlight key information, and the text echoes the data.

Next, I will use some examples to illustrate these principles.

Principle 1: Don't set the reading threshold too high.

I believe you look tired in the picture above. Charts are used to present the core idea of data, not to complicate it.

Principle 2: Don't have too many colors.

Seeing the picture above, my attention was completely attracted by the color.

Principle 3: Highlight key messages

If there are a lot of data, you can highlight the information that you want to highlight most, and you can mark the color as shown in the above picture.

Principle 4: Text and data echo.

This is a more complicated data chart above, and it is easy for people to lose sight of the key points. We need to interpret it and see the picture below for a better presentation.

In this picture, we can see that few people use X products, but most people are satisfied with X products and most people are not satisfied with X products.

These points are key information, so they are marked with colors on the map, and the text next to them is also drawn with corresponding colors to explain.

In short, all the expressions we used to talk about were words, but in today's data age, we need to learn not only words, but also numbers.

Data analysis forms insight

When we know the basic concept of data and can express it with data, we actually have basic data thinking.

But what we need to advance is how to form insights from data. Just like the story of Lin Biao, how to find the winning point from the data.

It takes only two steps to form an opinion through data analysis:

1) Find key data indicators

Every task and work we do must be measured by data indicators. Because, if you can't measure a thing, then you can't grow effectively.

For example, if you are HR, your important job is to reserve and retain excellent talents for the company. But if every time you report to your boss, you always talk about what tasks you have done, such as organizing an employee discussion today and organizing an employee run tomorrow, the other party will definitely be indifferent.

You need some indicators to measure, such as whether the turnover rate of key talents has decreased, such as whether the speed of talent recruitment has accelerated, such as whether people's efficiency has improved and so on.

For another example, we launch a new product and sell it to users. You want to know the feedback from users. At this time, some sales told you that the user feedback was very good and gave you a good screenshot, but some people said that some users complained that the product design was not good. Who do you listen to at this time? So, the key is to find those data indicators.

You can think about it, what data indicators should be used to measure your work tasks? As long as you find this, you have the basis of data analysis.

2) Master enough analytical methods.

Once you have the data indicators, you can analyze them. This is the most critical step for us to gain insight!

Common analysis methods include: classification analysis, matrix analysis, funnel analysis, correlation analysis, logical tree analysis, trend analysis, behavior trajectory analysis and so on.

Taking the work of HR as an example, I explain how to do the above analysis in order to gain insight.

0 1) classification analysis

For example, the brain drain rate is analyzed by different departments, different job levels and different age groups. For example, if you find that the turnover rate of a certain department is particularly high, then you can analyze it.

02) Matrix analysis

For example, if a company has an assessment of values and abilities, it can make a matrix diagram of the assessment results to find out the proportion of employees with strong value matching, employees with weak value matching and employees with weak value matching, so as to find out the health status of talents in the company.

03) Funnel analysis

For example, recording recruitment data, submitting resumes, passing the initial screening, passing through one side, two sides, passing through the last side, accepting the Offer, successfully joining the job, and passing the probation period is a complete recruitment funnel. From the data, we can see which link can be optimized.

04) Correlation analysis

For example, the brain drain rate of each branch of the company is quite different, so we can analyze the correlation between the brain drain rate of each branch and some characteristics of the branch (geographical location, salary level, welfare level, employee age, manager age, etc.). ) find the key factors that can best retain employees.

05) logical tree analysis

For example, if employee satisfaction is found to have declined recently, it will be dismantled. Satisfaction is related to salary, welfare, career development and working atmosphere, and then salary is divided into basic salary and bonus. In this way, it will be dismantled layer by layer to find out the changing factors in the influencing factors of satisfaction, so as to get insights.

06) Trend analysis

For example, the change trend of brain drain rate in the past 12 months.

07) Behavior trajectory analysis

For example, tracking the behavior track of a salesperson, from entry, to starting to produce performance, to rapid growth of performance, to fatigue period, to gradual stability.

This process can help managers to better judge the psychological rhythm of a salesperson, so as to better manage.

Here is a question to think about. If you are a salesperson, how should you use the above analysis method to look at the data?

To sum up, if you want to have data thinking, you can penetrate the essence of the problem from the data and use the data to influence others. The most important thing is:

1, have the basic literacy of data, including understanding basic statistical concepts and avoiding logical fallacies of data;

2. Be good at using data to communicate and express, including understanding the purpose and object of communication, choosing appropriate data expression types, and paying attention to data visualization;

3. Be able to find insights from data analysis, including finding key data indicators and mastering enough data analysis methods.