Many students who do data analysis have heard about the analysis model of people, goods and fields. However, this thing is a thing that only knows its name and does not see its real body. How to combine actual analysis? Today we will explain the system.
Problem scenario: A
fresh food e-commerce company has a low user repurchase rate. 60% of users have no second purchase behavior within 30 days. The operation leader is very anxious and requires data analysis to increase the repurchase rate. As a data analyst, what should you do? How to do it?
- Establish an artificial intelligence precision recommendation algorithm (collaborative filtering is used for 40% probability, and association analysis is used for 60%)
- Make a line chart of the repurchase rate at the beginning of the past 6 months, and then write three vigorous and powerful characters: "Go higher!"
- What is the analysis? Isn't being an e-commerce just giving out coupons? All users who have no repurchase will be given coupons, and you will be done with Aoli!
Or do you have other ways?
1. Cargo attribute analysis
Let me ask a simple question first: What is the difference between rice, white noodles, a barrel of oil and strawberries, cherries, and mangosteen? Even if you have never bought vegetables, you know that rice noodle oil is something that you have to eat every day, there is no seasonality; strawberries, cherries, and mangosteen are not eaten every day, and they are very seasonal. If you go to a vegetable market or a supermarket, you will know that the rice noodle oil is usually bought in whole packages or barrels. You can eat a barrel for a long time. There are also special rice barrels, rice boxes, and oil cans for packaging. Strawberry mangosteen are usually broken up for retail sale, and they are not durable enough to be put away.
These seemingly commonplace product knowledge are collectively referred to as: goods attributes. The attributes of the goods will directly affect the purchase behavior of consumers:
Purchasing frequency : Fresh vegetables and fruits are purchased frequently, but rice noodle oil is low
. Season : Fresh vegetables and fruits have seasonal products, but they are expensive and not delicious off-season. Rice noodle oil is not expensive. Seasonal
product prices : single product prices are expensive Sell less, buy cheap, buy in bulk, and buy cheap ones in bulk. Buying
channels : if there is logistics and distribution, it will be easier to buy large hard currency (rice noodle oil) online, and buy offline, best Can try some on the spot to avoid stepping on thunder
The attributes of these goods are common sense and laws of nature, and will not change due to the calculation of data indicators. Therefore, in fresh products, user behavior will be directly affected by the products purchased in the past-you can't count on a user who just bought 10 jin of rice, and then buys another 10 jin in two days. In other words, if there are users who come to buy rice over and over again, then you have to check whether the rice you provide is much cheaper than the market price. Someone is squeezing wool.
There is a simple matrix model that can describe the repurchase idea of fresh products. The core is the frequency of product purchases and product relevance. The frequency of purchase is explained above. Product relevance refers to certain products that are naturally bought together. Especially in the field of fresh food, such as buying frozen chicken wings and bamboo skewers, it is very likely to buy charcoal, meatballs, and barbecue juice. Therefore, the two-dimensional crossover has the following matrix (as shown below)
But note that only from the perspective of the properties of the goods, it is very incomplete. There are so many channels for grocery shopping, why users have to poke and poke in the app. Is the vegetable market not fragrant? What is the appeal of APP/WeChat Mall? This involves: the field problem
2. Analysis of store attributes
A quick question: What are you going to eat for lunch today? Don't think, answer right away!
Ten of the ten classmates couldn't answer, right. Actually make you hungry, you have to struggle ten or twenty minutes, let alone prepare in advance.
The same is true for buying vegetables. One very important reason why the elderly like to go shopping in the vegetable market is that cooking is not purposeful. You can buy something on the spot. Secondly, you can shop around and choose fresh and cheap ones. The visual impact of the vegetable market, including the fresh food area of the supermarket, is far stronger than that of e-commerce, which is the influence of the attributes of the store on the repurchase behavior.
Store attributes, including:
- Convenience: The closer and more convenient the market is, the more attractive it is
- Cleanliness: the cleaner the market, the more attractive
- Product richness: the richer the dishes, the more attractive the market
- Product freshness: the fresher the dishes, the more attractive they are
- Product price: Because of different shop rents and labor, some stores are expensive
In traditional offline stores, there is also a matrix model for the location of the store. (As shown below)
The indicators used in online channels are similar to those used offline, except that the user’s login scenario, login frequency, and content accessed after login replace the location of the store. Online channels can do much more analysis in terms of content and jump paths than offline channels.
What's interesting is that, unlike FMCG such as clothing, snacks, and toys, in the field of fresh food, the experience of online channels is worse than offline. Therefore, the advantage of online fresh food is reflected in the scene where you can't go out. For example, on rainy days, such as traffic control during the epidemic, such as commuting and no time to go to the vegetable market, etc.
However, this leads to the third problem: some users may simply want to be cheap, while some users really just need to buy online. Therefore, the human factor must be considered.
3. User attribute analysis
Note that in traditional industries, people are talking about people and goods, and people are salespeople, not consumers. The so-called human efficiency refers to the average economic benefits generated by salespersons. However, Internet applications are APPs to users, and there is no concept of sales. Therefore, salespersons are changed to users. The so-called analysis of people becomes an analysis of user attributes.
When it comes to user attributes, many students are conditioned to reflect: gender, age, and region. The question is that your company can really collect so much real user information? And these fields may not be able to see anything, the most typical is gender, the difference between male and female ratio is often only a few points, which can explain the problem.
Labels based on interaction and consumer behavior will be more useful. For example, in the field of fresh food e-commerce, how many customers register to send 20 yuan rice noodle fuel coupons, the first order is free of delivery fees, and imported cherries are 25 yuan 4 kg. of. This is called promotion-sensitive users. Similarly, you can also tag: rigid purchase users, abnormal weather purchase users, users in epidemic areas, etc. These tags may be more distinguishable (as shown below)
Fourth, people and goods yard model building
With a basic understanding of three dimensions, it can be used to comprehensively explain the problem. Back to the beginning of the "low repurchase rate of fresh food e-commerce" problem. You can first establish an analysis hypothesis from the perspective of people and goods yard:
Person perspective:
- The quality of the ground push is too poor, and the user has no demand
- The user has a demand, but there are too many wool types, and there is little rigidity
- There is a certain amount of users who just need it, but the product does not meet the needs of users
Cargo angle:
- The product itself has too few categories
- There are many categories, but no strong drainage models
- There are drainage models, but the price has no advantage
Field angle:
- User habits are not established, and secondary logins are rare
- Yes for the second login, but not to the purchase page
- Go to the purchase page, but no order
After establishing each hypothesis, there are two ways to establish the overall idea:
first, starting from the data, start wherever the problem is serious.
Second, starting from the business, what major events have recently occurred, and where to start
(as shown below)
Finally, you can twist the various analysis dimensions to form the overall analysis logic, and form a conclusion from coarse to fine (as shown in the figure below).
V. Summary
The reason why the three dimensions of people and goods are often used is that these three are directly related to user behavior, and there are some natural laws to follow in commodity attributes, store attributes, and user habits. Therefore, it is very suitable as a basis for analysis, deep and detailed. On the one hand, they can have a clearer understanding of the business; on the other hand, there are clues to build more complex models.
However, a common problem in the industry now is that newcomers in business know how to issue coupons. It is hard to say that a coupon is hidden. It is also called: Internet thinking is free! Newcomers who do data know about RFM, association analysis, and collaborative filtering when talking about models. Please, brothers, your platform users are sticky. Once you log in, users are born with coupons. How much real data do you have to train the model? Just like the fresh food e-commerce industry, I really go to the market a few times to talk to the main people who buy food: grandpa and aunt, housewives, it will be more useful than discussing AARRR with colleagues who are hungry every day. You can try it.
Some students will ask if there is a more general analysis scenario for the repurchase rate. If you are interested, follow the WeChat public account [Grounding Qi School]. In the next article, we will share the application of repurchase rate in the medical aesthetics industry, so stay tuned.
Author: Chen grounded gas, micro-channel public number: down to earth school. A data analyst with ten years of experience has launched a series of data analysis courses and has more than 20,000 students.