How to build user portraits for online education!

The implementation of the data-driven operation concept cannot just be limited to the blind construction of the system. It is also crucial to enable internal users of the enterprise to use it and make good use of it. Starting from this article, I will summarize and sort out how data empowers the marketing business line from the perspective of another marketing CRM product line that I am responsible for, combined with the marketing business scenarios of online education . This article starts with the core element of CRM - leads, covering an introduction to leads, lead life cycle, and portrait construction.

01. Introduction to clues

1. Definition of clues

A user lead is a collection of information used to describe a customer and represents a marketing or service opportunity. With leads, there is a quantitative basis for the development of sales activities. In business activities, leads are the most advanced sales resources. After circulation through CRM and sales follow-up, they are finally converted into customer resources. In order to facilitate sales colleagues to contact customers in a timely manner, mobile phone numbers or WeChat accounts are the core information of clues.

2. Source of clues

There are different standards for classifying the channel sources of leads. Based on my understanding of the sources of online education channels, I personally divide the leads into: external paid channels and self-built channels.

1) External payment channels

SEM: Carry out keyword placement and continuously optimize keyword placement strategies to attract users to the marketing campaign landing page or official website to collect user information.

Information flow advertising: Information flow advertising mainly uses data to locate users' long-term or potential needs and actively displays native ads to users. Common ones include Toutiao, Douyin, Kuaishou, Guangdiantong, etc.

Channel Agent:

• Online channel agents, provide leads and win commissions
• Offline institution cooperation, book QR codes, local training institutions, etc.

Resource purchase or exchange:

• Purchase whitelist data from data service providers
• Exchange resources with other companies to obtain leads

2) Self-built channels

Main site SEO optimization: optimize the internal and external content/structure of the website, improve the natural ranking of the company's website in search engines, obtain more exposure, and attract more target users to enter the main site and leave clues.

App traffic generation: Use app store operation and promotion, soft-text marketing and other means to increase APP exposure and guide users to download and register; in addition, for the online education industry, good operation of tool apps can bring it rich traffic.

Private traffic in the WeChat ecosystem: As the internet traffic dividend has bottomed out, the cost of acquiring customers has risen sharply, forcing traffic-dependent companies to change their thinking and shift their focus from traffic diversion to fission, allowing users to fission before purchasing; private traffic has three characteristics: owned, free, and repeatedly reached; the operation of private traffic in the WeChat ecosystem, through the combination of Moments, WeChat groups, official accounts, mini-programs, and corporate apps, forms a network for screening potential users layer by layer, and the connection and continuous optimization of enterprise micro-functions with individual micro-functions is conducive to achieving a win-win situation for individual micro-functions, enterprise micro-functions, and enterprises.

New media: Operate content accounts of company brands or personal IPs of celebrity teachers on platforms such as Douyin, Kuaishou, Weibo, and self-media public accounts, increase fans and user attention through offline or online activities, and trigger user forwarding and fission dissemination through activities.

3. Clue information composition

The information feedback from leads to sales staff should not only stop at informing them of marketing opportunities, but also provide high-value customer data combinations as much as possible, so as to help front-line sales colleagues understand users as comprehensively as possible, improve conversion efficiency, reduce their time in mining user information, and let them focus their energy on promoting single links. Based on the information richness of the clues, we can divide the information structure of the clues into three levels: low, medium, and high. The corresponding data sets are: user tag information, user feature information, and user portrait.

Obviously, the higher the lead information level, the more information the salesperson can grasp and understand about the customer, which is more conducive to closing a deal. Therefore, it is necessary to improve the lead information level. I will explain the construction of lead user portraits in Chapter 3.

02. Life cycle

The lead life cycle is the definition of the complete flow process of leads in the CRM system and is the core strategic logic of CRM system design. The design of the lead life cycle should fully consider the division and connection of various marketing and service stages, the actions of different employee roles at each stage, and the flow process and status of user information. Taking online education business as an example, the author gives an example of the life cycle of leads, as shown in the figure below.

• Lead generation: each channel interface reports leads to CRM, and the new leads generated must include channel information and tag information.
• Screening and cultivation: In order to ensure the integrity and validity of new lead data, the lead data needs to be cleaned, and deduplication and merging are required when necessary. The quality of the cleaned leads needs to be judged based on the scoring model. High-scoring leads should be promptly assigned to sales follow-up, and low-scoring leads need to be pre-cultivated in the user operation link to improve the quality of the leads and activate the allocation.
• After circulation and distribution, the leads that have been scored to be of follow-up quality need to be diverted based on the company's overall operating strategy, and the leads with different characteristics are distributed to the corresponding sales team lead library (departmental lead library). There are two ways to circulate leads from the lead library to sales: active sourcing and passive rule allocation.
• Follow up on leads and clarify the SOP process, sales follow-up words, and lead status. Taking the business situation of the author's company as an example, the status of each stage in the lead follow-up process is: pending follow-up, in contact, unclear intention, clear intention, appointment payment, and transaction.
• Transaction & Service: The first transaction of a new lead is defined as the first sale. The courses sold in the sales stage are generally paid trial courses or low-priced basic courses. The first-sale leads should enter the Do Not Disturb lock mode to prevent sales harassment from affecting the students' experience in the service link. In the middle and late stages of the service link, we must continue to explore users' intention to renew their reports and facilitate them.
• Deposition activation: leads that have not had clear intentions for a long time will enter the deposited inventory after the trigger time threshold (3 months or half a year). Users need to use activity operation strategies or recall activation strategies to guide users to generate purchase intentions, generate high-intention behaviors, and improve the lead quality score.

03. Portrait Construction

1. Portrait service construction

1) Introduction to the construction process

• [User label production]: Operation personnel create and maintain labels in the Sensors user portrait system through custom rules;
• [Integration of portrait modeling]: In most companies, due to the objective and individual conditions of business scenarios, user data (user-like data) is actually deposited in multiple systems, such as lead data in the CRM system, registered user data in the user center, delivery management platform data, community activity data, etc. In the process of building user portrait services, it is necessary to integrate scattered user data, regularly synchronize scattered user service data and user tag data to the data warehouse through the opening of data channels, and perform user portrait topic modeling in the data warehouse to ensure accuracy, completeness, and consistency.
• [Portrait query interface]: The big data platform provides a query interface based on the data warehouse user portrait theme, including but not limited to: basic tag information query, user full portrait information query, user tag value query, etc. The query interface provided should be based on the specific demand scenarios of the business system.

2) Sensors portrait label type

Customized labels can divide the population into multiple levels, and operators can customize the name and calculation rules of each level. For example: produce a label, divide users into three groups of low value, medium value and high value according to the set rules and conditions, and set the calculation rules for low value, medium value and high value respectively. The high value rule is: the user needs to complete ≥ 30 payment orders in the past 7 days and the total amount of goods in the payment order details in the past 7 days must be ≥ 6000.

Basic indicator values, such as the number of times a user completes an event, are used as label values. For example: produce a label to mark the total order amount of the customer's payment orders in the past 30 days. This label is a numeric label and uses the total amount of the user's payment orders in the past 30 days as the label value. If the user pays three orders of 100 yuan, 150 yuan, and 300 yuan respectively, then the total consumption amount of the user in the past 30 days is 550 yuan.

The first and last features use the time dimension or event attribute of the user's first or last completion of an event as the label value. For example, a tag is generated to mark the number of days since the user's first payment for an order. The tag is a numeric tag. If the user's first payment was made 2 days ago, it will be marked as 2.

Event preference label, taking the top attribute ranking of the event completed by the user as the label value. For example: produce a label to mark which product type is more preferred by the customer when the payment amount is higher than 300 yuan, and record the top 3 types that the customer likes most. This label is a collection type label. If a user has placed 20 product orders in the past 7 days, including 5 orders for maternal and child products, 4 orders for clothing products, 4 orders for fresh food products, 3 orders for mobile phones and digital products, 2 orders for daily necessities, and 2 orders for other products; then the customer's label will be marked as preferring maternal and child products, clothing products, and fresh food.

Behavior distribution results, for example: generate a tag to calculate the distribution of the number of days that users logged into the App in the past 30 days. This tag is a numerical tag. If the user has access records on the 1st to 5th day and the 20th to 30th day in the past 30 days, this tag will be marked as 16.

3) Tag data synchronization solution

Solution 1: Use JDBC + HDFS to export tags

Create a data table in text format and insert the data to be exported into this table.

CREATE TABLE default.age_export AS /*SA_BEGIN(test_project)*/ SELECT id, first_id, age, user_tag_*,user_group_* FROM users WHERE age IS NOT NULL LIMIT 10 /*SA_END*/

• Get the HDFS path of the data table: SHOW TABLE STATS default.age_export
• Just remotely copy the files in the above path

Solution 2: Sensors User Profile v1.5 provides a user list API

• Create groups/labels and complete calculations
• Call the crowd package file generation API to generate a file and obtain the file name
• Call the file download interface to download files

2. Establishment of clue label system

The establishment of a clue label system must rely on the characteristics of the products sold by the company, the user operation system, and the marketing characteristics. The establishment of a label system in an online education scenario can refer to the figure below, which starts from the three themes of value stratification, life cycle, and behavioral preferences.

3. CRM portrait presentation

The presentation of portraits is the 1/8 of the iceberg that emerges above the water in the entire portrait service system. Although the content is small, it is very critical. The quality of information presentation directly affects the efficiency of front-line sales. I will not go into details about the universal design principles. Here are some of the more important points:

• Structural and modular display of user information
• Similar information should be aggregated rather than dispersed as much as possible
• Display all image information within 1 screen or 1.5 screens
• Key information should be visually prominent

The above is all for sales personnel to obtain user information as quickly and accurately as possible and improve information acquisition efficiency. The following is an example of a simple version of the author’s first iteration, as shown in the figure below.

04. Final words

So far, this article starts with the core element of CRM, "leads", and gives an introduction to the leads, their life cycle, and portrait construction. In the lead life cycle section, the author mentioned the cultivation and activation of leads. In the next article, I will summarize the cultivation and activation of leads and the dynamic scoring model and provide corresponding solutions. Readers who are interested in CRM products and data-enabled marketing business please stay tuned!

Author: A data person’s private land

Source: A data person’s private land