The most complete analysis of user portraits, super detailed!

For Internet practitioners, a word is often mentioned - user portrait . As a data PM who is mainly engaged in user portrait DMP, I am always asked by the demand side at work:

I want to check the user portrait of XXX or whether it is possible to have a portrait of XXXX type of user. Or maybe other products will ask: How do you create user portraits?

However, during the communication process, I found that different people have very different understandings of user portraits. Some people think that user portraits contain detailed information about users, some think that user portraits can reflect the statistical characteristics of a group, and some think that user portraits can be used for user research... These ideas are more or less one-sided. This article explains the basic knowledge of user portraits, analyzes them in combination with some DMP products, and explains the application of user portraits in K12 products.

1. What are we talking about when we talk about DMP and user portraits?

This section combines common DMP & user portrait definitions with the definitions of DMP & user portraits in my work.

User portrait is a very important part of DMP, so we will explain DMP and user portrait separately.

1.1DMP

1.1.1 What is DMP?

DMP stands for data management system. Judging from the name alone, this definition is still very broad, so many domestic companies or individuals will misunderstand the core functions of DMP.

Based on my understanding, DMP is actually a comprehensive data collection, processing, and integration platform. It absorbs data from various data sources, takes users as the basic unit, cleans and organizes them into structured data tables, and calculates user tags in order to accurately describe various users.

A pure DMP platform refers to a small DMP technology service provider with strong customization capabilities and good neutrality. The US DMP market is extremely segmented, while the Chinese market is highly integrated. Often the demand for DMP is closely linked to DSP and SSP. Currently, it is difficult to have a pure DMP platform.

1.1.2 What can DMP do?

Precision marketing, advertising, personalized recommendations, and other applications

1.1.3 DMP infrastructure and data processing flow

DMP infrastructure:

DMP data processing flow:

1.1.5 Practical Application of DMP (Products Available on the Market)

DMP advertising platform: Tencent Guangdiantong, Alibaba Mama Dharmapan;

Independent third-party DMP: talkingdata, Sensors Data;

Personalized recommendations: Toutiao, Yidian Zixun, Taobao, JD.com, etc.

Note: We can feel the personalized recommendation application, but we cannot see the logic behind it.

Other applications

1.2 User Profile

1.2.1 What is User Portrait?

There are two types of definitions for user portraits: User Persona and User Profile

User Persona: is a typical user abstracted from the user group by product designers and operators. For example, during the user research phase, product managers use questionnaires and customer interviews to understand the commonalities and differences of users and summarize them into different virtual users;

Classic case - User portrait mentioned in "Elements of User Experience"

UserProfile: Based on each person’s user behavior data in the product, a collection of tags describing the user is generated. For example, guess whether the user is male or female, where he or she lives and works, which celebrity he or she likes, what he or she wants to buy, etc.

The user portrait mentioned in this article refers to User Profile

1.2.2 How to create a user portrait

Basic data collection: collect static and dynamic data of users on and off the website;

Behavior modeling: Based on the user's basic data, behavior modeling is carried out through technical means;

Build a portrait: Through behavioral modeling, a series of user tags can be output. Each user's tag can form a set, and this set of tags can represent the characteristics of this user.

1.2.3 Common Applications of User Portraits

Personalized recommendations (e-commerce, information products), risk control, forecasting, etc.

1.2.4 The Relationship between User Portrait and DMP

DMP is a data management platform, which can be simply understood as providing data to the DMP platform, and the DMP platform outputs a series of labels or other desired results.

User profiling is the process of inputting user data into DMP, which then outputs user labels.

DMP can not only output user tags, but also other tags. For example, it can input articles and output article tags.

Therefore, user profiling is one of the application directions of DMP.

2 Related product introduction

Since user portraits are mainly used as underlying applications, many of its applications are perceptible but invisible. For example, behind the personalized recommendation page of the e-commerce platform and the personalized recommendation on the homepage of the information app, it is the user portrait that plays a role, and the user tags and content tags/product tags are intelligently combined.

Since I am engaged in the K12 education industry, I only selected two types of competing products: open DMP platform and educational products, among which open DMP platform is the main one.

3 Competitive product analysis

3.1 Open DMP platform

3.1.1 Product Description

3.1.2 Function Comparison

Dharma Plate

GuangDianTong

Sensors Data

Note: Sensors Data looks more like a data analysis tool, but its underlying construction and data management are similar to DMP. The visible part is similar to BI functions, which can be regarded as the performance of DMP at the application layer. Therefore, it is also listed as a competing product.

3.1.3 Interface comparison

Note: Since all three products require payment to experience all product functions, the interface is mainly derived from the description document and may differ from the actual interface.

Dharma Plate-Tag

Dharma Plate-New Label

Dharmapan-Crowd Report

Dharma Plate-Crowd Details

Dharmapan-Overall Report

GuangDianTong operation interface

Guangdiantong-Create Advertisement

Guangdiantong-Create Advertisement 2

Sensors Data - User Analysis - Event Analysis

Sensors Data - User Analysis - User Attributes

3.1.4 Comparison of product underlying technical architecture

illustrate:

1) This part is based on the possible underlying architecture abstracted through product experience and reading documentation, and is not the actual situation;

2) Product underlying technical architecture: The functional modules we see are all realized by the collaboration of different technical modules. The underlying technical architecture of a product describes the underlying technical modules corresponding to the product and the relationships between the modules.

Dharma Plate

Sensors Data

3.1.5 Summary

In summary, Dharmapan and Guangdiantong are typical applications of DMP in Internet advertising. DMP is the core brain of targeted advertising. The user portrait provided by DMP is the most core and critical step in targeted advertising. The most important goal of Dharmapan and Guangdiantong is to show the right ads to the right people at the right time.

For Sensors Data, a type of data analysis tool, DMP also plays a very important role in the process of data analysis and data visualization, such as which types of user behaviors are more prominent and which types of users will have what kind of behaviors in the future.

3.2 Educational Products

Application of User Portrait DMP in Educational Products——

(1) Onion data personalized course formulation (similar to adaptive learning);

Before learning, users will take a test first, and a personalized course package will be prepared for the user based on the test results, as shown in Figure 1 below;

Figure 1

After the user completes the learning, a visual learning analysis report is displayed based on the user's learning test results, as shown in Figure 2

Figure 2

4 How to use user portraits?

User portrait is a low-level product, and the application of user portrait is usually difficult to see. So, how should user portraits be used? Combining the analysis of DMP products and educational products, the application of user portraits is summarized as follows:

(1) User tag visualization

Functions of related products: Guangdiantong and Dharmapan directly display user tags. Users can directly select tags and combine tags to select target users and then place advertisements.

Examples of scenarios that can be used as reference: For example, I want to send a push message to those students who have just signed up for the course this year, have high purchasing power and study hard, hoping that they can participate in a live broadcast to promote their mastery of knowledge. Then you can directly select the labels: new users + high spending power + serious study, and then send text messages to these users.

Benefits: Refined operations, improved ROI; simplified operations

(2) User tag association analysis:

Related product functions - Guangdiantong Lookalike: (1) Provide seed users; (2) Screen seed user characteristics; (3) Match seed users with Tencent users to expand the population

A scenario that can be used as a reference: A group of users purchased product A, and I want to find users who are more similar to those who purchased this product. Then you can import the information of this group of users into the DMP, calculate the tags of this group of users, and then find users who are more similar to this group of users through association analysis.

(3) Personalized recommendations:

Related product features: Personalized course development

Scenarios that can be used as reference (taking K12 education as an example):

(1) For unregistered users, we recommend personalized course pages to them based on their visit behavior to increase purchase conversion rate; if the users come in through Internet advertising, we can develop personalized landing pages for them to increase registration and purchase conversion rate;

(2) For those who have registered but not purchased a course, we recommend personalized course pages to them based on their browsing behavior to increase their purchase rate;

(3) Based on the user’s learning behavior (mainly the test-taking situation), develop personalized test questions and learning suggestions for the user.

I am a k12 education data PM, welcome everyone to communicate and grow together~

Author: Little Donkey Takes You Around the World

Source: Little Donkey takes you around the world