domeet webmaster

Students who do data analysis should have confidence in themselves.

Editor’s note: This article comes from the WeChat public account “Grounded Air School” (ID:gh_ff21afe83da7), author: Chen grounded gas.

Read this article for reference: Want to make a good user portrait? Learn this basic operation first

Some students asked: I see users blowing portraits everywhere, but I have never seen real examples. Today we come to a practical example of using user portraits to increase transaction volume. And this example happened around us. What Mr. Chen just met when he brought the baby last week is still fresh and warm.

1A live example of user portrait

Qiu Gao is refreshing. Dad went out to play with coco and saw a lot of people flying kites by a lake. Suddenly wonder: let’s go let it go! coco said: ah! So the two went to go ghost (Cantonese, refers to undocumented street vendors) to buy kites.

Seeing that Coco likes it, Dad is ready to pay for it. However, when the peak turns around, I did not expect to have another paragraph below

Then Dad and Coco fly a kite happily. But think about it afterwards: I have a big slot, this is not based on user portraits, use the recommendation system to improve the complete process of trading! Data collection-labeling-product recommendation-up-sell in one go, and also made a second recommendation, it is really clever to increase the transaction rate and customer unit price separately.

Although Uncle doesn’t have a big data system, and he doesn’t know how to write code, this idea of ​​doing things is far better than asking everywhere, “What are the algorithms for recommending kites?” “How do BAT sell kites?” I am in the Internet kite industry. Are there any newcomers who understand the Internet kite problem? After all, I have practiced my knowledge and copied IQ.

However, if you think about it carefully, things are not that simple: there are so many more toys, why are they kites?

2The real use of user portraits

Chang Liuwa’s parents are familiar with them. In the parks, ponds, and grasslands, there are small vendors selling bubble sticks, toy guns, magic wands, sand digging tools, kites and other toys. It seems that they are all selling toys, but you can take a closer look. The difference between different toys is very obvious.

01Toy Gun, Magic Wand:

This style of toy is very important! Whether it is the magic wand of Balala Little Magic Fairy, or the 98K that eats chicken, directly determines whether the children buy it. This is typical: explosion-driven. Explosions can be created artificially. The test here is business design and promotion ability. No money will be spent. No matter how good the recommendation system is, it won’t help.

02Bubble Stick:

The selling point of this kind of toys is very prominent: the sky is full of bubbles. So Ya does not need to recommend at all. You just need to ambush on the side of the road, wait for a group of children to come over, and then yell into the sky, a lot of bubbles flew towards the children. Then there are 9 out of 10 children who want to buy. This is typical: experience driven. The test here is the sales ability, so that users can experience the effect. Otherwise, if you don’t allow the experience, you will still be tempted to recommend it, but you will still lose it.

03Dug bucket:

The degree of homogenization of this gadget is very high (a square head shovel, a pointed shovel, aA rake, a spatula, a shell model, a starfish model, a small bucket, the same national model). And the function is very clear and single: dig sand! For parents, there is nothing to say, buy it when it’s cheap, and don’t buy it if it’s expensive, a typical price driver. Here is the pricing, the price is expensive, and then recommend, parents should not be wronged.

In comparison, kites are very unique:

• The kite is too big to be opened, it is not easy to display, and it is not easy for customers to choose.

• The color, size and graphics of kites are indeed a lot of people will care about. You can’t help but pick people out.

• The unit price of kite is relatively high, and it is expensive to sell. The adult turns around and walks away in minutes, selling cheaply and not making money.

Especially for Uncle Walking Ghost, if the kites are all spread on the ground, it is estimated that the city management will not be able to run away! In addition, dozens of kites are spread on the ground, the appearance of the kites is also very poor, and the selection of parents is also very difficult. You know, no parents are willing to take their children to the stalls to buy things. Most of them are cautious ways to fear/being crying and making trouble. Therefore, if you add a little more choice burden, you will lose more chances for customers. Therefore, the recommendation system works well here.

Comparing these four categories, we found that the recommendation system is just an auxiliary tool for commodity management, which is suitable for non-explosive products with certain differences between categories. However, every tactic of merchandise management is related to the user’s portrait. Only by understanding the user can we accurately meet the needs. So don’t think of the recommendation as soon as you mention the user portrait, there is more work to be done (as shown below)

3Why is it difficult to make useful user portraits!

Since the user portrait is so easy to use. So why do we always think that the user portrait is a bunch of data piled up there, is it useless in the end? Because we want to make useful user portraits, we need to avoid too many pits.

Pit Point OneNot clear what to do

If Uncle Ghost meets his parents, noAsk people what to buy and chat in a chat. Can he still sell kites? Of course not, maybe it’s cool to talk nonsense. This is the difference between targeted and non-targeted. Corresponding to the work, many people make user portraits based on: “Leaders demand to do” “I think everyone else is doing”

As for:

• What did you do?

• Which department does it use?

• In what scene?

• What indicators are improved?

• What is the current indicator?

• How much is expected to increase?

• What packages are needed to improve the indicators?

Not at all thought

If this works, it’s damned. ╮(╯▽╰)╭

Pit Point TwoData quality is not guaranteed

Note that the uncle clearly saw a dad than + a child coming, but he still asked: “Adult or child?” Without taking it for granted, this is a professional performance. Because toys are a typical scene where users and buyers are separated, especially kites, adults and children can play. At this time it is important to confirm the authenticity.

Corresponding to work, it is what we often say about data quality. Data quality is the premise of all analysis, and many companies underestimate the rigor of data collection. After the so-called “label diffusion method” came out, a group of newcomers who did data thought they didn’t want to collect it. As long as there is an algorithm to calculate the real data, this is really a self-cutting way. Data quality is always better.

Pit Point ThreeDo not label but only count

Note that the uncle asked for a label, not a raw data. For example, adults may have a height of 150, and children may also have a height of 150. So why ask the label instead of a specific value? First of all, it is difficult to collect data from tags. Secondly, this tag not only represents height, but also represents aesthetics, and its role is much richer than the original data.

This is a visual expression of the role of the label: rich in meaning and easy to use. This is why the data has to be collected and the labeling must continue. Labels are refined, meaningful data classifications that are more useful than raw dataMuch.

Pit Point FourThe effect is not verified, and there is no iteration

Speaking of the usefulness of labels, some people have suffered from label fanaticism and mad marking. No matter how useful or useless it is, how useful it is. In short, the more labels, the better. But there is no label with verified effect, as there is no. Not to mention the more complex secondary and tertiary labels based on primary labels.

Interestingly, Uncle Ghost uses a secondary labeling strategy. Note that if you say at the beginning: ordinary kite 20, children kite 30, it is very likely that the parent will choose 20 directly. If at the beginning it says: kite 30, it may scare parents away.

But first confirm that the parents are willing to buy, and then recommend an expensive reel that does not hurt your hands. The success rate is greatly improved. Because it has been confirmed that this parent is willing to accommodate the child, and the parent who is willing to accommodate the child will definitely buy an expensive one. In fact, according to incomplete statistics, half of the scenes are reel that doesn’t hurt your hands. Hey, the taste is your taste, your taste.

4Summary

Of course, the above is just an exercise by Mr. Chen. The uncle may just be smart enough to think so much. But this intuitive example is very suitable for everyone to remember. When you are confused next time, “What’s the use of user portraits?????”, you can think about it again.

There is still a small problem here. It seems that this process is very simple. Why do you need a professional data analyst to do it? Business people can summarize it themselves. Answer: There are two reasons. First, the actual business scenario has a large amount of data and many dimensions. It is very time-consuming to process and requires professionals to do it. If you see that your operations are not designing schemes, choosing gifts, and surveying users, but are doing data 8 hours a day, then the company is not far from bankruptcy.

Secondly, the experience of business personnel is easy to be faced by short-term effects, and then make wrong judgments (as shown in the figure below). The KPIs of business personnel are in charge, and it is easy to choose the short-term effects and ignore the long-term effects. At this time, data analysts are needed to calm down, long-term observation, and experience. In order to better guide the business.

So, doStudents of data analysis should have confidence in themselves. The value of data is more than a complex model, basic work, method precipitation, experience summary, and feedback business, all we can do. The method is not as difficult as possible, but as useful as possible. Remember to remember.