What should data analysts do when they are always faced with various difficulties?

Students who work with data may encounter many difficulties. So how to solve this problem?

Students who work in data science, have you ever been challenged at work? What are some common problems that make people hate you? From the topics that students complained about before, I have selected 8 high-frequency problems. Let's take a look at them today. Attention! High energy ahead, prepare your blood pressure medication~

Question 1: "It's just a number, why does it take so long?"

This problem is very common. Most people don’t know what “a number” actually means, and the first thing children learn in kindergarten is counting, so people mistakenly believe that “getting a number” is as easy as counting.

To solve the problem, we must first educate the business about where the data comes from, and that it does not fall from the sky. We must also establish a set of standards and persistently educate the business:

1. What needs to be re-collected?

2. What has been collected but not cleaned and is unusable?

3. Which ones are available, but the query conditions are very complex

4. Which ones are available and easy to query?

5. What do not need to be queried and have fixed report output?

When promoting, promote at the same time: if you want a quick response, sit down and discuss the needs and fix them into reports. This can not only improve business awareness, but also promote our BI system, increase the utilization rate of BI, and eliminate scattered and temporary needs, which is the long-term solution.

Attention! When promoting, don’t post articles like “Tencent/Byte/Huawei’s Big Data Applications” that describe how powerful big data is, which will only whet the appetite of the business. You should forward more books and skill trees on big data/data warehouse/data governance/data analysis, so that they can directly experience “Wow, this thing is so complicated!”

Question 2: “Our data is huge and it’s all there, why can’t you analyze it?”

This question and the previous one are brothers and sisters. The essence of both is that leaders do not understand data and think that a few numbers are "big data."

There is another side of this problem. When you are recruited, your leader says earnestly: "Our data is huge and it's all there. All we need is for you to analyze it." You must be careful at this time, because their data is probably not cleaned up, and the product codes and order numbers are very messy transaction data.

If you add the clause “there is no data team” or “you are isolated from the data team”, please be cautious when joining the company, as you will be PUA’d to the point of doubting your life.

Question 3: “Isn’t data analysis a matter of data analysis? Why should I be involved?”

Note that this sentence is a question, not a rhetorical question. The business may not really know what the relationship is between it and the data. At this time, we need to patiently guide:

1. Collect data first before you have data

2. Data collection requires business processes

3. User basic information must be collected

4. The page needs to be embedded to have user behavior data

5. Product information must be maintained in a standardized manner, otherwise we cannot understand it.

6. Supplier information, material information, activity information, and coupon information are the same as above

7. If you use them randomly during business activities, you will not be able to tell which data is correct.

…

In short, explain the pros and cons in detail and try to win business support. When you are thirsty, the business still ignores you. Remember to write meeting minutes. Afterwards, the data is a mess. Someone should take the blame...

If this sentence is a rhetorical question, please refer to the next one for the solution.

Question 4: "Just do your numbers, why are you asking so many questions?"

There are two forms of this problem. The first form is that the business comes to us to get data, but cannot clearly explain the caliber of the data. At this time, no matter how aggressive the other party's attitude is, you must confirm it clearly in advance. If you don't confirm it clearly in advance, he will still blame you afterwards: "The data you get is not accurate! Everything you do is wrong!" If the other party is really aggressive and stupid, then use the data dictionary to let him check which indicators to look at and how long to look at.

The second form is that we want to take the initiative to understand the business afterwards, but end up getting rejected. Note! Some people in the business department are difficult to please, so don't force all business people to cooperate. If they really don't want to communicate, just deal with them in a businesslike manner and find someone else to communicate with. Generally, there are one or two departments that cooperate well, and the other departments can just deal with it.

Question 5: "I already knew what you did. Do you have any in-depth analysis?"

This question is a sister to the previous one. They are both caused by poor business cooperation attitude. If you don't tell me what you know, how can I know whether you know it or not? As long as we have actively communicated, the responsibility is not on our side. If we have not communicated, then we should reflect on it and communicate more actively until we are met with nothing.

In fact, there is a win-win solution here: the business gives the analysis hypothesis in advance, and the data analysis verifies the hypothesis afterwards. Not only can it directly address the issues that the business cares about most, but it can also help the data understand the business's thinking, and finally a win-win situation. However, this requires the business department to have a sense of cooperation, and everyone can choose people to act.

Question 6: “How accurate are your predictions?”

This question is always asked. Be careful! Don't argue with him about "80%, 90%, or 95% which one is more accurate". This will trap you. If there are any problems with the business later, he will blame you, "Your prediction is inaccurate. My product gross profit is only 5 points. What's the point of your 95% prediction? It's all your fault!"

The essence of this question is: What should the business do with the forecast? !

• If the natural trend is predicted first and the business then makes judgments based on the natural trend, the final result will be interfered with by the business behavior and will definitely be different from the prediction.
• If we consider business behavior first and then predict the results, the business must first tell us what they are going to do, otherwise the prediction will definitely be inaccurate.
• If the predicted indicators have nothing to do with the business, then the poor performance of the business should not be blamed on the prediction (it has been said that it has nothing to do with it, but you still want to pass the buck!)

Of course, the ideal situation is: it is predicted that there will be problems, but after business efforts there will be no problems. The business then praises the data analysis for timely discovering the problems, and the data analysis praises the business actions for effectively stopping the problems. Wouldn't it be nice to praise each other? Why hurt each other?

Question 7: “If you can predict 100% accurately, I will definitely be able to do a good job in business!”

There are many similar answers to this question, such as "I didn't do well, it's all because the data analysis couldn't pinpoint the problem 100% accurately." In short, if the business is not done well, the blame lies with the data, and the business itself is not responsible.

Many students will ask back: "Why is my analysis inaccurate?" This is a trap. Don't argue with these people.

The details of the analysis, in the end, he will always blame you, such as: "Why can't you analyze with 100% accuracy that the user will open the APP on which day and which second!" Unless we have a magic crystal ball, this is impossible.

The best way is to use the power of others to defeat them. Attention! Bosses hate this kind of business that shirks responsibility. So a better way to deal with it is to do a layered analysis and summarize more examples of successful business. With the same data analysis support, why can others do well, but why can't you do it well. Direct the spearhead to "why you can't compare with others" and "how to do the business". This will not only avoid shirking responsibility, but also solve the problem down to earth.

Question 8: "How do you prove that your analysis is related to the company's performance improvement?"

This question usually comes up during performance appraisals. When you hear this question, you will be so angry that you want to say, "You were like a dog when you were looking for numbers, but now you think I'm ugly after seeing the data!" However, it is usually the leaders who ask this question. Apart from resigning in anger, is there a better way?

In fact, all support functions will encounter this kind of problem to some extent. In essence, they are "tool man" functions. They are useful only when they are used, and forgotten when they are not used. To solve the problem, we must start from the source of performance evaluation. Not only must we have reasonable evaluation standards, but we must also deliberately create more "beautiful moments" so that leaders can see and remember them.

Author: Down-to-earth Teacher Chen; Source public account: Down-to-earth Teacher Chen (ID: 773891)