Click "Read Original" for details -01- This is a technical article. Based on past experience, technical articles often have low readership. Currently, the infrastructure for marketing digitalization is not difficult to complete, but application is difficult, and the difficulty lies in the fact that applied technology is a stumbling block. Marketing digitalization can only achieve precise, real-time interaction through user profiles. Therefore, the digital business logic must be expressed using internet technology, and this hurdle must be overcome. Precision means user profiles. Real-time means instant profiling and updating, pushing information and policies within a time imperceptible to users, and engaging them. Precision and real-time are not achieved through people, but through technology. The reaction time of technology can be so fast that users do not perceive it. -02- Basic understanding of big data business: Consumer behavior is predictable, and the basis for prediction is not the usual demographic classification, such as gender, income, etc., but consumption behavior data. This differs from traditional marketing logic, which is basically based on demographic classification. The basic logic of big data business is: Consumer labeling. Labels are the consumer's DNA. DNA originates from consumption behavior, has a tendency, but is also changing. Labeling is a simplification; simplicity is dimensionality reduction, and dimensionality reduction is to stay within the limits of human cognitive capacity. User profiles, from the perspectives of brand owners, agents, and retailers (platform operators), each have different profiling bases. First, the perspectives differ; second, the goals differ; finally, the relationships differ. -03- First, look at Alibaba's big data concepts and applications. The following content is selected from "Disruptive Marketing: The Business Revolution of the Big Data Era". The author Chen Jiehao provided big data consulting to Alibaba, and co-author Che Pinjue is former Alibaba vice president and chairman of the Data Committee. I. Concept: Finding customers does not look at demographic attributes, but at behavioral labels Traditional marketing tends to handle heterogeneity issues, classifying and segmenting based on demographic characteristics such as gender, age, residence, income, etc. Gender is a demographic indicator; personalized labels are behavior and motivation. Demographic indicators conform to statistics. Personalized labels can predict behavior and accurately predict the next purchase time. Behavioral labels do not look at demographic attributes, only at purchase behavior, such as dividing consumers into three categories: new customers, main customers, and dormant customers, and then tracking them by category. Behavioral labels only look at recordable behavior (behavior with data). For example, the first online purchase, regardless of how many offline purchases, the online record is the first. II. Model: Big data marketing enters the new 4Ps from the traditional 4Ps Big data 4Ps: People, Performance, Process, Predict.

  1. People model: New customers (N), existing customers (main customers E0, dozing customers S1, half-asleep customers S2, dormant customers S3).
  2. Performance model: Increase in customer count, increase in average transaction value, and increase in activity.
  3. Process: Improve the revenue equation through a hierarchical execution process.
  4. Predict: Predict the time of customer's next purchase, allowing merchants to talk to the most likely customers at the right time. Through control, real-time observation, zero-lag communication, and personalized information can be achieved. III. Goal: Predict purchase time and tendency, improve revenue Based on big data, the next purchase time can be predicted. Recommendations should be made before the predicted purchase time, recommending products that match the user's purchase tendency. The goal is to improve revenue. IV. Big data simplifies labels Personalization cannot mean one label per person. Alibaba simplifies up to 1000 demographic labels into 6 dynamic groups and 19 labels. Customer dynamics (NES): (1) New customers; (2) Main customers; (3) Recent returning customers; (4) Dozing customers; (5) Half-asleep customers; (6) Dormant customers. Tenure (Length): (7) Early; (8) Mid; (9) Late. Recency: (10) Recent buyers; (11) Mid-term non-buyers; (12) Long-term non-buyers. Frequency: (13) High-frequency buyers; (14) Medium-frequency buyers; (15) Low-frequency buyers. Monetary: (16) High-spending buyers; (17) Medium-spending buyers; (18) Low-spending buyers. Next purchase prediction (NPT): (19) The most likely time point for repurchase within 7 days. -04- The biggest differences between brand owner digitalization and retailer digitalization are: First, retailers have unlimited SKUs, while brands have limited SKUs. Second, a difference derived from the first: Retailers have many substitute products, while brand owners have no substitute products. Third, a difference derived from the second: Brand owners' digitalization pursues brand repurchase (loyalty to the brand), while retailers pursue platform stickiness; retailers provide associated sales, while brand owners only have repurchase. Fourth, retailers have "product profiles" and "user profiles", while brand owners may not need product profiles. Fifth, retailers' digitalization requires "product-user matching" to improve conversion rates; brand owners' digitalization pursues cognitive conversion and repurchase rates. -05- For platforms or retail, the main purpose of user profiles is "product-user matching" to improve conversion rates. In addition to predicting the next purchase time, an important aspect is to improve the match between products and users, i.e., the timing and characteristics of recommended products match the user. Brand owners' user profiles have three important goals: First, repurchase, i.e., repeat purchase, turning new users into old users, and old users into big users. Because brand owners have fewer product options, repurchase is a must. Therefore, user behavior profiles, along with policies matched to the profiles, are key to inducing repurchase. Second, activating the B-end. Using the C-end to activate the B-end is an important goal of brand owners' marketing digitalization, and it is a requirement to use small data to activate the big market. Third, user fission. Although Pinduoduo also has user fission, because brand owners have fewer online options, user fission becomes more important. -06- In addition to the important differences in big data profiling goals mentioned above, compared with brand owners' user profiles, platform or retailer user profiles also have the following differences: First, differences in user profiles:
  5. Retailers only have C-end user profiles, while brand owners also have B-end user profiles.
  6. C-end user profiles are mainly based on user attributes, while B-end user profiles are mainly based on behavior.
  7. C-end user profiles are pushed automatically in real-time, while B-end user profiles can be used for interpersonal communication. Second, policies targeted at user profiles. Based on user profiles, formulating targeted policies is necessary to lock in users and cultivate big users. Retailers' policies for users include platform policies and merchant policies. Brand owners' policies for C-end users include brand owner policies and agent policies. Third, online push and post-push online. Platforms or retailers have many SKUs, forming a product ecosystem, with high user online frequency and long duration. Therefore, they can push products or policies in advance based on predicted next purchase time. Brand owners have limited products and may have no other product options. Therefore, the system should push policies in advance to induce users to go online. Or during the online process, push policies at the right time to induce orders. Fourth, platforms or retailers can make associated recommendations based on user profiles because they have too many SKUs. For example, they can push a series of products to mothers, but if it is a milk powder brand owner, they can only cultivate repurchase or become a big customer. Fifth, platforms or retailers need to profile both products and users simultaneously, and then match them. Brand owners basically do not have product profiling and matching issues during the sales process. Sixth, brand owners' C-end users are only for the C-end. Brand owners' C-end user profiles must also be linked to the B-end, using the C-end to drive the B-end.