How to become a mid-level data analyst

On the road of data analysis, many novices are often confused about how to advance. This article will reveal the three keys to data analyst advancement and provide specific guidance on how to implement these strategies to help data analysts take a solid step in their careers and achieve a win-win situation of personal growth and business value.

"What exactly is considered advanced?" is a question that many new students working in data science often ask in Knowledge Planet.

There are so many articles on the Internet about "how to get started" and "quick start". But after I actually started to do data analysis, I realized that what I learned was just bullshit. I was just running numbers every day, but the real algorithm work was far away from me. So what is the future?

The core of the above doubts is that most of the articles on the Internet are based on textbooks. Because the first chapter of the textbook talks about pycharm and anaconda, the second chapter talks about pandas, the third chapter talks about matplotlib, the fourth chapter talks about numpy, and the fifth chapter talks about sklearn.

So the authors take it for granted that doing data analysis is the first step of installing the software, the second step of collecting data, the third step of visualization, the fourth step of machine learning models, and the fifth step of the business kneeling at their feet and saying: data analysis is awesome, come and drive me...

Wake up!

After you actually enter a company, you will find that the biggest problem with data analysis is that it has little presence. Data drives the business? That is: the boss uses data to drive the business . As for the data workers, in most companies, they are just miscellaneous people. Those product managers, operations, sales, and planners all think that they understand analysis, and they just lack the ability to run numbers. As for the data workers, just focus on running numbers.

Moreover, those who work in business especially like to say: "I know how to do data analysis after reading the data analysis articles on WeChat Moments, but the data analysts in my company can't even do such a simple thing as accurate recommendation with big data. It's all their fault!" - People who work with data not only have miscellaneous work, but are also easily blamed.
Therefore, if you have really worked in a company, you will understand that if you want to get out of a predicament, you really need to do three things:

1. Strive for opportunities for independent projects, not to be a toilet cleaner
2. Be clear about the scope and boundaries of work, and don’t take the blame for no reason
3. Show work results and achievements, and strive for internal recognition

With these three, you can make more contributions and take less blame. These three points are the real transformation from a novice who waits for others to count to a mature data analyst who can stand on his own. How to do it specifically, let's briefly talk about it below.

1. How to win independent projects

During the learning stage, newcomers have practiced so-called "projects" on the Internet, such as Titanic, Taobao shopping, and a certain country's credit card. The biggest difference between real corporate projects and these Internet celebrity projects is that no one arranges what to do for you . If you sit and wait for others to arrange it, you are just waiting for a phone call: "The boss wants this data, and it must be given before the end of the work day."

If you want to win a project, you must do a good job of regular data demand statistics. If you want to find opportunities in trivial, scattered, and daily work, you need to rely on careful analysis, not charity from others (as shown below).

Based on the demand statistics table, we can proactively discover:

• Which departments have greater demand?
• Which needs are urgent?
• Which ones started from 0 to 1?
• Which ones require complex analytical support?
• Which ones can be undertaken with fixed tools?
• What are the boss’s concerns?

This way, you can avoid the embarrassment of asking business questions when you encounter problems, and then being rejected with a "none of your business". Finding departments willing to cooperate and finding valuable cooperation projects depends on this little bit of accumulation, not falling from the sky.

2. How to clarify work boundaries

The purpose of setting clear work boundaries is to avoid being blamed, there is no other reason.

We must keep in mind the Three Purities Principles:

1. It is important to distinguish whether there is data or not!
2. It is important to distinguish whether there are standards or not!
3. It is important to distinguish whether you have a plan or not!

Without data, there is no analysis . This is nonsense, but it is the easiest to ignore. In the eyes of the business, it is always: "Our data is huge, and it is all there, we just need someone to analyze it" - a big pile of shit is still big . The business will not deduct the details of the data. If you don't do your homework in advance, you will face a pile of dirty data afterwards, and you can't cook without rice.

Without standards, there is no way to evaluate . This is also nonsense. However, the sales staff often say: "I just want to increase sales; I just want to improve activity; this is what the boss said to do, why do you care so much?" It seems to have a goal, but it is not specific. If you don't take the initiative to remind them at this time, they will want to supplement it later. It will become: if you say the business is good, you are just a yes-man and meaningless to the boss; if you say the business is not good, you are waiting to be criticized by the sales staff. It is a dilemma.

Without a plan, the forecast is inaccurate . This is also nonsense. Can the conversion rate of betting on a 100 yuan coupon be the same as that of betting on a 10 yuan coupon? Can the conversion rate of a good copywriting be the same as that of a bad copywriting? It is all nonsense to talk about forecasting and trends without considering the business plan. However, the books that teach forecasting all talk about data processing methods, and rarely talk about how to combine them with reality, so this step is often forgotten.

The above three principles are the key points summarized by many newcomers after many setbacks. However, they are also the points that newcomers are most likely to forget. Because in the self-study stage, they practice with ready-made data sets, ready-made backgrounds, and ready-made books. No one has ever taught them how to solve specific problems and communicate in specific ways. This makes it easy for problems to occur when working.

3. How to show your work achievements

The results of data analysis are similar to the story of "Columbus standing an egg on its head" - before you say it, everyone thinks it's impossible; after you say it, everyone says: I thought of it a long time ago! This is very simple . So simply reporting a few numbers or making a few suggestions verbally cannot prove that this is your achievement. Instead, it makes the business smarter, and the future analysis needs will become more and more complex and difficult to handle.

Therefore, there are three standards for data results:

1. Output quantity can be quantified
2. Results can be reused
3. I don't understand process encapsulation

The details are shown in the following figure:

If you want to achieve these three standards, you can’t just rely on writing PPT and making oral reports. It is imperative to develop data products. But it is unrealistic to upgrade from scattered data directly to a complete data product - the business can’t wait that long and won’t stop its daily work. Therefore, you must be aware of product upgrades and gradually transition to complete products.

The technical capabilities we have accumulated are used on this occasion.

Within the permitted range of time and data quality:

1. If you can make reports, you don’t need to obtain data temporarily.

2. Those that can be uploaded to the system do not need manual reporting

4. If you can use the model, you don’t need business rules

5. If you can solidify the rules, don’t run the numbers every time.

6. Those that can solidify the standards do not require special analysis

In short, with each upgrade, manual operations, temporary operations, and personalized operations will be reduced. The more product functions are enriched, the higher the prediction accuracy, the faster the query speed, and the simpler the problem location method, the greater our value will be.

The above is the idea of ​​breaking through from novice to intermediate. When is the cultivation considered successful?

To put it simply, it means to be independent .

• In terms of results, mid-level data analysts can solve problems independently.
• At work, a mid-level data analyst can handle the “dumb questions” of the business.
• In terms of methods, intermediate data analysts can accumulate experience instead of copying from everywhere.

The specific performance is: when you are interviewing or reporting at the end of the year, you don’t need to say silly: I did a lot of analysis. Instead, when you can clearly talk about the amount of work you have done, the output products, and the analysis model, you will be considered successful.

Unfortunately, many newcomers do not pay attention to these issues.

Rather than having in-depth discussions on specific issues, they prefer to grumble about how their company is too low-end and that if they could work in a big company like Tou Teng A, they would definitely be surrounded by beautiful scenery, chirping birds and fragrant flowers.

Rather than thinking deeply about business scenarios and business processes, they prefer to look at the "underlying logic" and "core model", and are tirelessly searching for the "National Authoritative Certification Method" on the Internet.

Rather than solving problems, they prefer to have a copy of "Master Machine Learning in 21 Days from Zero Basics", thinking that if they learn this, one of the companies will be impressed by them - in short, there are too many complaints and too few details, so it will be difficult to make progress.

After being able to face and solve problems independently, we can discuss what skills a senior data analyst needs to lead a department of more than 10 people. In terms of results, senior data analysts must not only solve problems, but also understand "what to achieve" and actively guide business development.

At work, this article only discusses what happens when the business makes stupid mistakes. There is another kind of situation: "Not stupid, but bad!" Senior data analysts have the ability to influence decision-making, so they have to face more bad guys and have the ability to fight wits and courage.

Author: Down-to-earth Teacher Chen

WeChat public account: Down-to-earth Teacher Chen