Decoding The Current E-Commerce Pattern: How Does JD Use Data To Achieve Precision Marketing?

A few days ago, JD.COM Li Xingyi, the business ecological director of the POP Platform Department Data operation The thought war behind "was the theme of his speech at the Beijing Yibang Madman's Conference, in which he made a speech from technology, marketing, logistics Merchants explained in detail how JD uses data to achieve precision marketing and how to use data through application scenarios from these five aspects.

According to the information, Li Xingyi mentioned that it is very important to predict sales, because if the sales volume is not good, a large part of the inventory will be formed, which will be a big cost. The core of JD's self support is product selection, pricing and stock up. The advantages of JD's self support are its own production and marketing and product management. It has a good grasp of this product. In many cases, it has entered into in-depth cooperation with manufacturers at the product design stage.

In addition, Li Xingyi also introduced JD's data products for businesses, including the data compass, which has been updating its products and functions. This part of the function is particularly powerful. JD Easo, which focuses only on search, also has a campaign plan, which is specialized in precision marketing, and some advertising products, such as Precision Connect, belong to advertising products, It is used by the advertising department for businesses. These applications can help businesses and JD self run businesses very well.

It is understood that the other three guests at the meeting made in-depth analysis from the perspective of "how traditional enterprises use data operation", "618 data: decrypting the current e-commerce pattern", and "how state-owned retail brands take the road of big data".

The following is the transcript of Li Xingyi's speech:

I mainly want to share with you some of JD's dry goods in the 618 process, so that you can have an in-depth exchange. Today's sharing is five pieces. JD's big data applications are many in all aspects, How does JD use big data and how does it use this data through application scenarios.

First, take a look at JD 618. If a friend goes to the hall on the first floor, he can see a large hall. There are many large screens on it, one of which is dedicated to monitoring all aspects of 618. This screenshot is used to predict sales. For example, you can see the gray background when 618 started, He will predict the sales and traffic of JD as a whole, and make some predictions on various core indicators. The red one is the actual data, and you can see that the prediction is very accurate, and the actual is very consistent with the prediction, including some of the peak points we said, for example, 618 was a peak at the beginning, and then everyone went to work, about nine o'clock, There will be a peak. This prediction is very difficult, including some of my friends who are self-employed. They clearly know that it is very difficult to make this sales forecast. There are many things to do. It has many factors, including the need to consider the sales in the past, and the current traffic situation. You are like JD and Amazon, In this industry, the sales forecast is relatively accurate, which is very helpful to stock up, because if you do self support, if the sales forecast is not good, a large part of the inventory will be formed, which will be a big cost. Of course, if you have the opportunity to go to Jingdong Hall, you will see the whole 618 screen, It will also monitor all links, including the order link, the delivery link, and the after-sales link.

As for technology, we will all do a risk control, which is also very core, including our merchants and security. We will monitor all aspects, including some malicious orders, whether users are a normal order, including some golden beans, or some risks, You think JD should have more than 100000 SKOs on its orders. If SKOs are hundreds of millions of SKOs, how can we monitor and standardize more goods? For example, whether some stores of goods conform to the rules of the platform, and whether the words in the pictures conform to the requirements of the advertising law. When goods are uploaded in real time, we will monitor them, In particular, we know that the background color, style and format of pictures are required for different categories of pictures, including whether it is yellow skin ringworm, and whether the text of activities and the text of activities conform to the advertising law when many activities are carried out, If there is a problem with the picture or text of this product, we will force it off the shelf, or even for more serious products, we will send some notices to the stores to notify them to pay attention to these situations. We have been doing this through the system, because if you want us to do it manually, a billion products and hundreds of people will be a problem for decades.

Purchasing is also part of JD's self support. The core of JD's self support is three parts. The first is selection, the second is pricing, and the third is stocking.

In terms of product selection, the difference between JD's proprietary product technology and other e-commerce platforms is that JD's proprietary product technology and production and marketing are very familiar with this product, because our brand dealers will have a very deep cooperation. For example, this game book, we will have product technology for each variety to connect with brand dealers, and will tell them, When selecting products, the product will have an in-depth communication with the brand at the initial stage, telling him the appearance of the brand, such as the appearance of the notebook, the attributes of the notebook, and some parameters of the notebook. We will carry out in-depth cooperation with the neighboring brand manufacturers based on historical sales, and help them to design the brand. For example, It was called Gamebook, which was popular in JD before. Gamebook was launched on this platform, but there was no such concept before. At that time, we used some data analysis and some search behavior of buyers to find a special group. He bought this notebook just to play games. There were a group of young people, and they had very high requirements for each configuration, In order to play the game smoothly and happily, we found a brand of game book to cooperate with. We first launched the concept of game book, and then there will be a special one on the home page of JD. That picture is called game book, promoting the game book. Later, game book was promoted on the platform of JD. Later, there will be many e-commerce platforms, Our offline businesses have all launched the concept of game book. There will be many examples of this. JD's own advantages are its own production and marketing and product management. It has a good grasp of this product. Most of the time, it has entered into in-depth cooperation with manufacturers in the product design stage.

The second one is pricing. We are responsible for the technical research and development of the retail platform. It will give a reference price to the historical sales volume of this product and to the production and marketing. It will also compare some reference prices with e-commerce brands and give a reference price. A development trend of JD in the future is that most of its own commodities can be automatically priced and automatically adjusted at any time.

The third is to stock up. How to stock up this product is also very particular. This Lenovo two in one tablet was launched in 618 this year. When you search for tablet computers on JD, Apple's tablet computers are basically at the top of the list. Apple's brand is very strong, but we also saw the in-depth cooperation between JD and Lenovo, Lenovo's two in one tablet computer also sold particularly well in 618 this year, with more than 10000 units sold in the whole store, which was also the result of activities during 618. This one is the proprietary product of JD. Lenovo will launch a lot of molds before the notebook is sold or launched in JD, and then conduct in-depth discussions with our production and marketing, The production and marketing will give him a good suggestion. This tablet computer did very well during the 618 period, which shows that we want brand e-commerce to have a deep cooperation with brand companies. Now we can search for tablet laptops on JD, and Lenovo is very ahead of others.

Then there is marketing. No matter whether it is a business or a brand, marketing is inseparable from marketing. The most important part of marketing is precision marketing. The concept of precision marketing is easy to talk about, but if you really do it or if it is effective, it is also very difficult. How can I find precision users and how can I paint users, This involves a lot of previous modeling and user labels, which is not easy to do. JD has a product called the Ranger Plan, which talks about many of JD's digital products. It is based on different applications to solve different problems. This Ranger Plan is a data analysis product for users' precision marketing.

Users on JD can be divided into four parts, one of which is For users, let's take Handu Yishe as an example. The brand on the market is very good, but there are still some users on the Internet who do not know this brand, but they may be potential users of Handu Yishe. We define these users as blue ocean users, and the second is potential users, which means that they search or introduce through friends, It's the first time we learned about this product, or we browsed the brand's products. This is called a potential user. A new user is the first one to place an order. Existing users are old users. Many products have been well understood. We classify these users differently and adopt different strategies for different users, for example, For a new user, it is to enable the new user to place orders on the brand continuously so that he can survive. For an old user, it is to solve the problem of shopping rate. Don't let him fall asleep. Don't let him not buy for half a year, right? It is also to wake up his re purchase rate. It solves different problems.

We will run different departments or brands to analyze users from these dimensions. For this model, we have found JD's own products, and about two or three hundred have used this product, including Inman, who is also our user, and will give different targeted coupons to different users, After this was sent to him, we found that the conversion rate of most businesses who place orders through coupons is very high, which is much higher than that of those who do not issue coupons, compared with the average level of JD, or compared with the industry level, but some businesses, such as home furnishings, have a particularly high expansion price. If we issue coupons to place orders for them, The influence is still small, but the revival rate of most merchants whose expansion price is less than 500, such as clothing, shoes and bags, is very high, so different marketing strategies or the same marketing strategy have different effects on different categories and stores. Therefore, we should develop our own marketing strategies according to the characteristics of our own stores and brands.

The second is user portraits. User portraits are not only what we say e-commerce platforms are doing. There are many businesses, including listed companies that we have been exposed to. They are also managing their own members. In establishing their own membership system, JD's user portraits will be based on several pieces. One is that we will be based on the basic attributes of users, including age, The basic attributes of gender are also purchasing power, its expanded value, what stage of the expanded price is purchased, and user behavior characteristics. He Ping is a behavior characteristic of browsing websites, whether he searches for keywords or through insiders, or directly comes to JD website through some brands. Based on his browsing behavior, Whether coupons are sensitive or not is classified as well as some hobbies. Users have hundreds of tags and more than 200 tags. We will recognize these users based on these tags. User portraits will have many application scenarios, including personalized services. As I will talk about later, there are also many aspects, such as robots, JD will automatically have some people to answer your questions, and these will have some applications. This is what we call a smart store. When a user comes to the home page of JD, he sees this page, and The activity page, as well as the product page he is interested in, are different. They are automatically recommended based on the user's tag and user's portrait. They will recommend different products and activities to different users.

I don't know whether other e-commerce platforms will do it. JD has a product called traffic visualization. For example, we can do 618 or any activity. We can't see any activity. Our insiders will see that there is such an analysis platform, real-time data, and every inside story, Whether the implementation effect of each activity page meets our expectations, there will be a monitoring data in this section. If we find that the product does not meet our expectations, we will replace the active product.

Therefore, this analysis platform is very helpful to the implementation effect of our analysis activities, and can also improve the GMV and conversion rate of the entire platform. This product has been used on JD for a long time. It is used internally by purchase and marketing personnel, including operators. This is actually an application of real-time data, that is, how to monitor the effect of activities in real time. As long as you buy something on any page of JD, you can implement real-time data monitoring. Many products and screenshots are shared with you for the first time. {page_break}

In the area of logistics, JD is just talking about the commodity portraits and user portraits. In fact, JD is still making another portraits, called community portraits, community portraits. We will think that community users will have some common characteristics, because there are differences between communities. We have done an analysis of a community before, and we found Peking University and Tsinghua University respectively, Shangdi Community and Wangjing Community found that there were different bookstores in these communities. Peking University bought a lot of books from Chinese academic societies. There were also many books on traditional science, physics and mathematics. Tsinghua University usually bought books on engineering. Shangdi Community usually bought books on IT. It should be because there were many such enterprises, including Baidu IBM, But if you look at Wangjing, you will find a very interesting phenomenon. Users of Wangjing will buy books about Judaism and psychology. At that time, we didn't know why. Later, we asked some friends who live in Wangjing. They said that there are many Judaism restaurants in Wangjing, and there are many high-level people who might like to read psychology. This is a phenomenon, However, it is not easy to explain the reason. This shows that different cells have common characteristics. In fact, we also found that Wangjing and Tiantongyuan bought this kind of electric cooker. The expansion price and brand of this kind of small household appliances bought by Wangjing will be higher, while Tiantongyuan will be lower.

Therefore, based on the characteristics of different communities, JD has many applications, including in some communities, in our own cabinets, we will put some goods in advance. If someone wants to buy something in a hurry in the morning, they can buy it themselves, which is much more convenient than the convenience store in the community.

The second is the mobile store. I take JD's iPhone 6 as an example. This activity is also very relevant. It was the first time that it was launched in JD. At that time, what we asked was to deliver the goods to users as soon as possible. At that time, we had many mobile warehouses in each community. One by one, we would prepare the goods in that community in advance, Once the user placed an order and delivered the goods, we immediately delivered the goods. The fastest order at that time was delivered 12 minutes after the user placed the order. We delivered the goods immediately through the mobile warehouse. This application is also based on user portraits and community portraits, which is to arrange the goods in advance. We even thought that in addition to the mobile warehouse, Could you also put the goods in the distribution point, and put them in some warehouses near the distribution point, so that I can quickly send them to the buyer. In this way, the user experience will also be improved very well. The logistics, the mobile warehouse in the community, and the layout of the spots mentioned in this section are all based on the portraits of the users behind and the portraits of the community. Of course, we should do more accurately, and the data will be more, We need to dig deeper into the user's behavior. We have a lot of data to cooperate with Tencent, including the data of our community, and will also cooperate with many third-party partners. Of course, more is based on JD's full chain data. JD delivers goods directly to the door, and the delivery staff has a strong grip on users, The delivery staff can see the characteristics, age and shopping habits of the user when they look at the package. Their description is very detailed.

The last one is merchants. There are also a lot of data products for merchants, including the data compass we speak of, the data products JD provides for merchants. In addition to the data compass, there are the previous Ease of Search, which only focuses on search, as well as the canvassing plan, which specializes in precision marketing, and some advertising products, such as Precision Connect, which belong to advertising products, It is used by the advertising department for businesses. If the data compass is used, it has been updating products and functions. This part is particularly powerful. The store diagnosis was launched in March and April this year. Its original intention is to help businesses understand whether their stores are healthy. When I come to the data compass, I know whether I am healthy. I don't need to look at the data analysis for a long time, Just tell me directly whether there is a problem with my store. If there is a problem, what is the problem.

We analyze from four dimensions, including the flow of our stores, the conversion rate of our stores, the sales volume of our stores, and the decoration of our store pages. Each indicator should have its own data, and it will also be compared with businesses in the same industry. After comparison, we will see what is worse, whether it is good or bad, and whether there is a comparison, Then he knows that he has made some optimization on those core data, so he can know the conversion rate. If the click rate is not high, there are many ways to improve. First, you need to know what level you are at compared with the average level of the same industry. If there is a problem, we will mark it in red. If there is no problem, It is normal that we will use green marks.

This function is particularly helpful to businesses. We find many businesses to use it and get a lot of feedback. The following is the product of Easo. At that time, we did a lot of research, including some market analysis, and found that many businesses do not know how to do JD's SEO, nor what JD's search rules are. Can you tell me whether my SEO is effective? Many businesses' products are based on the weaknesses of businesses, This is also a product that we intended to do last year and launched this year. It is open to all users. This has a great advantage. Through JD's own distribution, you can get first-hand real-time data information when searching your own keywords and insider rankings. You know those keywords are well done, We can focus on optimizing some core keywords, and then also divide real-time data and front-line data. We know that SEO has black hat and white hat, so it can normally guide businesses to optimize product titles and SEO. It actually has many search rules, which can help businesses and JD self support. If you have done operation If you use it, you will have a deep feeling about how to master some search details, especially the accurate advertising, which is also very helpful for keywords.