"Have you ever done advanced data analysis?" As soon as this question came up, I asked many students. Ma yeah, I usually run to get count orders. I haven't seen any advanced data analysis before, so I answered. The system will answer today.
1. Popular explanation, what is considered advanced
Ask a simple question: Can I drive if I navigate the car? Answer: Absolutely. In fact, navigation has not been popular for a few years. But without navigation, driving will be extremely troublesome: can't find the road; miss a big circle at the intersection; stupidly stuck in the road and won't go around... In short, driving efficiency is much lower. At this time, only the old drivers who keep in mind the various road conditions can reach the end quickly and accurately-this is what people usually think of as a "senior driver". If you interview him, he must have many "advanced driving methods" to share.
But with navigation, the cost of driving learning is greatly reduced. Before a rookie couldn't find the road, now you can follow the navigation to reach the destination. Although the advanced drivers will definitely be a little faster, the advanced level has been greatly reduced-because the gap in results has narrowed a lot. Although the old drivers still have many complicated and difficult-to-learn skills, the results are similar, so that people no longer believe in them. Instead, they began to complain about their various vices: plugging, compacting lines, changing lanes without lighting...
So we see that the so-called advanced has two kinds of understanding:
1. Advanced in business: Mastered by a few experts, the result is fast and accurate, and the oral is cool and complex.
2. Advanced technology: It can help a large number of rookies, improve efficiency as a result, and operate simply and easily.
So here comes the question: Is data analysis a technology? Or business?
2. What is needed for advanced data analysis
The example of navigation is because data analysis and navigation are very similar:
So in theory, the most advanced data analysis results should be similar to navigation:
Really advanced data analysis is a systematic operation, with business processes as the guarantee, data collection as the basis, reports as the backbone, data products as the selling point, and business experience precipitation and model assistance. It is a set of simple and easy to use Tool system (as shown in the figure below).
However, if during interviews or foreign exchanges, some people who do not know how to do it often come out and mutter: You are not advanced enough. why? ? ?
3. Why are there so many people who don’t know the goods?
The more advanced the data analysis, the easier it is in the eyes of the rookie!
Because there are too many down-to-earth parts of it, are
they not cool, fantasy, and tall enough?
They will keep yelling:
The data is huge!
- Isn't it just a report?
- Isn't it just a reminder!
- Your prediction is too simple!
- Can I not say it, you automatically predict what I think!
If you try to explain to them: this just seems simple, you need to open up n systems, do n more buried points, collect n more data, and perform n repeated experiments. Just like you have to explain to him that the navigation software needs satellite remote sensing, real street shooting, and pre-calculated paths-he neither understands nor thinks it is very advanced. They will continue to yell: Isn’t that navigation can be done by many people? Isn’t it just inputting an address? What's the difficulty? In short, for them, simple operation means simple method, as long as they understand the name, it is equivalent to understanding the process. What they desire is awesome stuff that the process is incomprehensible and the effect is unexpected.
It is Di, what the rookies need is not a data analyst, but a wizard. Wearing a pointed hat, holding a magic wand, and wearing a gray robe, he said: Avada Krafla! Then it turned into a pile of banknotes. If you don't open your mouth, he will know the horoscope of the donor today-how advanced it looks!
Of course, there are still people who know the goods in the industry, but if you encounter this kind of rookie, you still prefer to be more honest with you: "Do you have any advanced methods", how should I deal with it?
Fourth, how to improve the sense of advanced data analysis
Let's take the seemingly simplest sales analysis as an example. Note that the following methods are only suitable for facing bad guys who are not knowledgeable and arrogant. In essence, this question comes from the incomprehension of data analysis and the over-confidence of one's own ability. So if you want to go back and stop the opponent's spirit, you can do this:
The first step: anti-customer-oriented. Take the initiative to say what he wants to attack you. Go his way, leaving him nowhere to go.
Step 2: Show the miracle. Note: To evaluate whether the data analysis method is advanced, the essence depends on the effect. So I want to talk about an advanced thing, first of all, what are the benefits of doing this. (As shown below)
The third step: Quoting the classics. In essence, novices like: model, thinking, paradigm, such a huge name, so it is a good name. For example: "I used data analysis to find five dimensions that are highly correlated with sales performance", which is directly called "building a sales force model." Did you force the grid up? Similar: "I conducted a cluster analysis of sales according to 5 dimensions and divided them into 5 groups" directly called "building a hierarchical and accurate operation system"... It's absolutely good!
The fourth step: blooming flowers. Don't explain too many details of the operation. If you explain too much, he understands it and thinks that you are not "advanced" enough. Similar: "I extracted the sales list according to the XXX rules and handed it to the business department for follow-up. After one month's inspection, I found that 65% of the predictions were correct and 30% had errors." It was too down-to-earth. Directly called: "Establish an enabling system, carry out 5 iterations, and continuously optimize model performance" directly put people on the ground.
After almost a few steps, the other party was either exhausted or blown out to satisfaction. If there is a sincere cooperation, we will talk directly. If it is someone who deliberately finds faults, there is no way to speak up-because he himself doesn't know where to go. As long as those people who use high-level methods every day, when they encounter data collection, data cleaning, and landing procedures, they will basically be ashes. You can't resist even if you want to resist.
Every time Mr. Chen goes to meet similar high-profile customers, he likes to download their APP directly or go to their stores. The core is concerned with their data collection process and activity rule setting. When I switched to the WeChat account to gather newcomer wool over and over again, let me hear from the sales/shopping guide: "Sir, just fill in this as you please", I will take screenshots and record recordings. Later, when I was talking about various advanced, intelligent, and magical methods, I would throw out these basic data quality problems for everyone to see, and then the topic basically turned to: Can skyscrapers be built on top of the cesspool, hahaha. The effect is crowded.
Of course, as practitioners, we still hope that the impetuous and blind atmosphere in the industry will be less, and everyone will work more seriously. This is also the reason why Teacher Chen works hard to popularize science. And some tasks here, such as predicting performance, such as predicting response rate, still need to use certain algorithms, which are more technical than running reports directly.
If you are interested, pay attention to the down-to-earth teacher Chen, let's share in the next article: How to make a really useful model. Stay tuned.
Author: Chen grounded gas, micro-channel public number: down to earth school. A data analyst with ten years of experience and CRM experience in multiple industries.