Daily dry goods and good articles to share丨Please click + follow
Welcome to follow Tianshan Intelligent WeChat official account, we are a vertical community focusing on the fields of business intelligence BI, big data, and data analysis.
Students who are interested in business intelligence BI, big data analysis and mining, machine learning, python, R and other data fields, add WeChat: tstoutiao, invite you to enter the headline data enthusiast exchange group, data enthusiasts are all here.
Note: This article was first published in Tianshan Community on March 29, please contact the author
if you need to reprint it Why just look at the average-
Xiao Wang is an operation specialist of a hotel, he wants to count n reports for the operation and marketing department to see. There is a table called "Members' Consumption Status Table". It is necessary to count the consumption status of various members. For example, we have 100,000 diamond members and consumed 20,000 last month. Each person has 2 orders, each order is 220 yuan. Similar to this kind of data.
However, Xiao Wang is very curious. Why only look at the average? Idiots know that diamond members consume more than gold card members, and gold card members consume more than ordinary cards. What if you break down more? ? He curiously raised a demand, and ran the consumption situation of all members in the past year, sorted by the total annual consumption, and divided it into 10 equal parts. A strange thing happened.
actually ranked the bottom 10% of the members, each order is only 100 yuan, more than 60% lower than our hotel rack rate! By the way, when did our house have such a cheap room? When has the marketing department done such a big promotion? Who are these people, and why do they consume at such a strong promotion? Xiao Wang was full of curiosity. For each of these people at , the company will lose more than 100 yuan. When all of them are counted, the company will lose more than 15 million yuan. Obediently, a lot of money seems to have the motivation to continue research.
-Curiosity is the driving force of analysis-
Curiosity 1: Has anyone deliberately shed wool? From the data point of view, these junk members only consume 1.05 times on average, that is to say, some people consume more than 2 times. Then, with such a low discount, they consume more than twice. It is very likely that they are deliberately collecting wool, such as collecting n orders. A single set of premium membership card offers. Therefore, the first step is to distinguish the card level among junk users and shave out high-end membership cards. Sure enough, under analysis, nearly 10,000 wool cards were found.
curious 2: What event gave such a powerful coupon? Among the remaining members of , since they can't get benefits from high-end cards, they must be coupons given by the event. Then what event will have such a powerful coupon? The second step is to check back all the promotion activities of the marketing department and the business department in the past year, see when, where, and what channels the coupons were distributed, and focus on whether there are coupons that can be used on top of the rules. Look back at these junk customers. The result is that there are two types of typical scenes, one is the new store diversion, with greater intensity; the other is the use of multiple coupons, which does not occur when the new store diversion.
curious 3: New store drainage is of strategic significance. What kind of ghosts are usually coming out? Xiao Wang did not rashly throw the analysis results up, because it is a politically correct thing to divert new stores, and even if it loses money, the marketing department may want to do it. He set his sights on those non-new store users, and further understood: Who are these people? After analysis, these people register earlier than the time of consumption, and only consume when they have coupons. It is very likely that they are long-term stays and other discounts, and they only consume old fried dough sticks when they look at the benefits.
is curious 4: Since these stackable coupons are so problematic, why has no one noticed it before? However, Xiao Wang did not immediately present this result, because since these coupons have such a big problem and have been taken away in large numbers, why has no one noticed it before? Doesn't the marketing department have to summarize every activity? He carefully turned back to the previous report and found that when junk customers concentrated on using coupons, the evaluation of the promotion was based on the effect of the entire event, how many active and how many customers were attracted by the entire event, no one could see where these coupons went. , High-value customers are also using it. "Fortunately, I didn't send it out hastily, otherwise it would definitely be sprayed by the marketing department, sayingThe cancellation of coupons will affect the consumption of high-value users. "
curious 5: What if I don’t send it? , if I don’t send it, high-value users won’t be able to consume it? With this question, Xiao Wang starts the fifth step: look at high-value users Consumption habits. The results found that high-value users are mainly male, business users, and the proportion of repeated consumption within 60 days is very high, and the consumption rhythm is very strong, and they will not consume according to the rhythm of issuing coupons (the only time used is just to make time) . Then you can boldly assume that even if there is no coupon, these people will continue to consume.
So far, the situation of junk users is very clear:
1. There are users of scalp wool, and
must be restricted from the rules. Differentiating past consumption behaviors will attract a large number of old fritters arbitrage.
3. The reward method for high-value users can be further upgraded to reduce hard placement of coupons and reduce costs. After comprehensive calculations,
optimizes the placement of coupons and cuts down the wool users. The company’s net cost savings are more than 10 million. Xiao Wang won the leadership’s award and was promoted to the director of operations analysis in the second year of his employment. (*^__^*)
——What is the bloody truth——
The bloody truth is: This problem is not the found by Xiao Wang, an operations specialist with a monthly salary of 5! There is no Xiao Wang with a monthly salary of 5,000 to make an analysis and countermeasures, and no Xiao Wang with a monthly salary of 5,000 is promoted and raised. This analysis is an annual salary. It was made by Mr. Dong, a professional consultant of 400,000. In fact, before Mr. Dong took over the project, there were several Xiaowang, Xiao Zhang, and Xiao Li in the customer's operation department. They have read this report for several years, and no one When a problem was discovered, no one raised the problem, and no one thought about solving the problem. The
old Dong and Teacher Chen mentioned this matter, which made Teacher Chen sigh. This analysis process is too simple to be simple, just add, subtract, multiply and divide and compare groups. It’s okay; it doesn’t use very deep knowledge. An 800-word article can explain it clearly, and you can understand it even if you don’t know how to do it. However, why the commissioners, Xiao Wang, Xiao Zhang, and Xiao Li would give up With such a good opportunity, do you have to wait for the company to spend a lot of money to ask a consultant to solve the problem?
With this curiosity, Mr. Chen asked Mr. Dong for some data. After desensitization, he made a few questions and asked verbally. Some cousins found the following pits, which greatly hindered them from going deeper:
pit 1: No curiosity. They will not take the initiative to look at the details. The 70% commissioners died here and will not take the initiative Thinking about !
Pit 2: The business department is sensitive. For example, when you see a consumption of 100 yuan per time, there are few people who think that the consumption is low. It is not possible to contact "This is only half of the regular house price. When did the marketing department put such a large Powerful coupons". About 15% of the people are pitted here.
Pit 3: I won’t make assumptions. I don’t think there are actually at least three games here. view. The data was rushed out, and the leader was led: "It may be this way, it may be that way, although it is low, it may also have tactical use." This place probably pits 10% of the people.
Pit 4: No reverse verification. Since the problem is so big, why no one has discovered it before; most of the commissioners did not think of this sentence, and lacked the idea of reverse verification for a problem. About 5% of the people are pitted here. The commissioners were wiped out here.
Pit 5: Will not ask the possibility. What would happen if it were removed? After doing this step, you can make your own analysis tenable. The positive statement: "It's not cost-effective to do it" plus reverse verification "It's not bad if you don't do it." The combination of the two can make your analysis very solid. , Will not be prevaricated by people for several reasons.
If Xiao Wang can find this kind of problem, he is also a Lao Wang-by Lao Dong
Yes, Xiao Wangs generally don't think seriously about the data at hand. When they have time, they will add various operation discussion groups on the Internet, ask people for book lists, complain about boring work, ask if they want to learn R, python or something-by Teacher Chen
The funny thing is, in Teacher Chen Among the 30 or so commissioners tested, 20+ are dissatisfied with their current jobs. At least 15 are really asking if Mr. Chen wants to learn R or python... Well, what Mr. Chen wants to say is, even if everyone improves Once you have the ability to code, you still have to learn to think actively, analyze data, andIt is just as important as analysis. Only one number is counted. It is still difficult to get a promotion if it fails to meet the requirements of analysis.
Students who are interested in business intelligence BI, big data analysis and mining, machine learning, python, R and other data fields, add WeChat: tstoutiao, invite you to join the headline data enthusiast exchange group, data enthusiasts are all here.
This article comes from the blog of teacher Chen Wen in Tianshan Community, and it is not allowed to be reproduced without permission. Original link of
: https://ask.hellobi.com/blog/chenwen/7036.
