Click "Read Original" for details -01- Digitalization without user profiling is just a decoration. User profiling is tagging users. Tagging is for easy identification, and it is done automatically by computers. Only with precise identification can we serve users precisely. Give a pillow when someone is sleepy. Automatically identify users, automatically push instant information and policies, and automatically form a new user profile the moment a user changes shopping behavior. The power of digitalization can only be demonstrated through user profiling. -02- In the past, offline sales were "one-to-one" and "face-to-face," whether for B2B or B2C. One-to-one allows relatively ample time for communication and interaction. If it's an acquaintance, recommendations can be based on preferences. In KA stores, users have to choose for themselves. Now users are online, and we need to face a massive number of users simultaneously. For example, during Double 11, hundreds of millions of users are online at the same time, millions of merchants are online, and hundreds of millions of SKUs are online. Moreover, instant push and instant updates are required. So, how can we quickly generate products or content that meet user needs with every click, without the user feeling any time lag? The answer is user profiling done automatically by computers. The platform automatically matches based on product profiles and user profiles, completes it instantly, and pushes it immediately. However, the user profiles on various internet platforms are done by the platforms, and users and merchants are unaware. For FMCG manufacturers, marketing digitalization cannot rely on platforms; they must rely on the brand owner to complete user profiling themselves. -03- User profiling is a technical term. Ordinary people don't understand it, and there's no need to understand it. Even if users have been profiled, they are unaware. In fact, as soon as you open various e-commerce platforms, Douyin, or Toutiao, users have already been profiled by the platform. Otherwise, how could the platform push products or content so accurately? Whether it's a transaction platform or a content platform, the platform is a matching system, matching products with users, recommending products to the most suitable users at the right time. Matching requires precise alignment, and what is pushed is exactly what the user needs. Only with matching will the conversion rate be high. On Alibaba's platform, with hundreds of millions of users and hundreds of millions of SKUs, without matching product and user profiles for instant push, wouldn't the matching process be chaotic? Toutiao and Douyin have massive content and massive users. If users had to search and find things themselves, what a waste of time. Fortunately, AI technology is now used for user profiling! User profiling is based on the traces (data) left by users' browsing, viewing, and transactions on the platform, according to certain profiling rules and purposes, to profile users, and then push content precisely based on the user profile. For example, media like Toutiao and Douyin push content based on user preferences and tendencies. When the US suppresses TikTok, China promptly issues policies prohibiting the export of algorithm-based technologies. One important role of AI algorithms is user profiling. AI algorithms are used both for user profiling and product (content) profiling, and then achieve matching between the two. For example, the platform determines a user's needs and preferences, such as price preference and brand preference, through user profiling. At the same time, products also have profiles, and then matching between product and user preferences is achieved. Only with matching will the conversion rate increase. The platform has already profiled products and users, but why don't users feel it? This is because user profiling is done by the platform, and merchants and users are only profiled; the profiles are used to push products or content. Users only feel that products are more to their taste, but they don't know they have been profiled. -04- User profiling on 2C platforms is already routine for the platform. So, does the digitalization of FMCG manufacturers need user profiling? Of course! Only with user profiling can it be more precise. For example, when a new product is launched, in the past, deep distribution adopted full-scale distribution. But now, with product upgrades, new product distribution needs to be precise. If a high-end new product is launched, it is necessary to find precise distribution terminals, so terminals need to be profiled. Suppose the terminals for high-end new product distribution must meet three criteria: 1. The terminal has the ability to recommend new products; 2. The terminal has a high-end user group; **3. The terminal has an advantage in that category. Then, terminals can be profiled according to the above criteria, and those meeting the conditions can be selected for cognitive education (such as experience), and then distribution. Because the profiling is precise, if sales are good after distribution, a more comprehensive distribution can be carried out. So, for new product distribution, where does the data for terminal profiling come from? There are two major sources: first, if you have your own terminal data, you can profile terminals based on historical data; second, if you are just starting and have no data, you can find a system platform or third-party professional company, and they will profile terminals based on data from other companies. Whether it's a manufacturer, distributor, or new retail, as long as you are online and digitalized, you must have user profiling. In current new retail scenarios, even offline, user profiling can be achieved. For example, when a user walks in a store, at a certain spot, a special screen will recommend suitable products based on the user profile. Another example: as long as a chain store recognizes the user entering (such as facial recognition), it can proactively guide shopping based on the user profile, which is more precise than passive guidance in the past. -05- Even traditional sales have simple user profiling, such as the cashier keyboard at 7-Eleven convenience stores, which is used to collect statistics by gender and age. However, this kind of profiling can only be collective, such as gender ratio and age ratio, not precise individual user profiles. User profiling is roughly divided into two types: one is user attribute profiling; the other is user behavior profiling. User attribute profiling includes gender, age, income, interests, active time, and residence. The high-end new product distribution terminals mentioned earlier also belong to attribute profiling for B2B users. User attribute profiling can be used for product development, such as finding target users; it can be used for product recommendation, such as if the user profile is "mother," then based on the needs of mothers, recommend suitable products to "mother" users. Online recommendation systems call users with the same profile "neighbors," and recommend to users based on the preferences of "neighbors." Traditional marketing also does user profiling, but more of it is attribute profiling. Big data also does attribute profiling, such as for new product R&D or B2B user profiling. However, big data profiling for C-end users is more about behavior profiling. If attribute profiling is about "guessing" user behavior based on the profile, then behavior profiling is about predicting the next behavior based on past and current behavior. Since behavior has already occurred, predicting the next behavior is relatively easier. The most important profiling in marketing digitalization is behavior profiling. Behavior profiling is different from attribute profiling; attribute profiling has a certain stability because gender and age are stable, and interests are relatively stable. However, even for the same person, behavior changes greatly. For example, on Douyin, the content a user likes today might be different tomorrow; they might want a change. So, as soon as the user makes a change, the behavior profile changes immediately. Many Toutiao users have complained that Toutiao has solidified their preferences, and they actually want to see more new things. But even if they "want to see," as long as they don't act on it, the profile won't change. As soon as the user tries to change, the profile changes immediately. So, what is user behavior? On Toutiao, user behavior includes clicking content, reading time, likes, comments, etc. Profiling is done based on these behaviors, and then the content to be pushed later is determined. Amazon is the pioneer in using user profiling for recommendations. Amazon uses user behavior on the site, including browsing items, purchasing items, adding to cart and wish list, as well as ratings and other user feedback, to form a user profile, and uses it for the following purposes: 1. Daily recommendations. Based on the user's recent browsing and purchase history, combined with current popular items, a comprehensive recommendation is given. 2. New product recommendations. A content-based push mechanism is adopted to recommend new products to users. Since new products have less user preference data, content-based push solves this problem. 3. Related recommendations. Data mining techniques are used to analyze user purchase behavior and find item sets that are often purchased together or by the same person. In book purchases, this type of recommendation is very common. 4. Others' purchases/browsing items. This is collaborative filtering recommendation of items. Through social mechanisms, users can more easily find products they are interested in. -06- User profiling is the soul of digitalization. It is hard to imagine that digitalization would be done well without user profiling. Traditional marketing also has numbers, but they are all statistics, such as annual sales and daily sales. These numbers are useful, but they are of little value for online operations. User online presence raises a requirement: instant recommendation. Every online behavior of the user, including browsing, shopping, rating feedback, group buying, etc., is changing the user profile. That is to say, user behavior profiling is dynamic. Each online behavior is followed by the next behavior, and between the two behaviors, it is necessary to decide what product or content to push next. Therefore, in an online environment, user profiling must be able to achieve instant recommendation. Instant online recommendation requires that user profiling be based on original data, complete profile modeling (build a model), complete calculations instantly, and then push products, content, or policies. The page updates after each click on platforms like Douyin are the result of user profile calculations; each click on Alibaba or Pinduoduo is the result of instant user profile calculation and push. For FMCG manufacturers' digitalization, currently it is basically only about being online; there is no modeling, no profiling, and no instant push capability. For example, with one code per product, most are just indiscriminately sending red packets. Without user profiling, users are treated as the same profile (the same appearance). User profiling is a big topic and will be discussed repeatedly. Next issue will discuss the difference between user profiling from the perspective of brand owners and retailers.