In-depth case study: Accurately outline user portraits and build an efficient financial customer analysis system from 0 to 1

With the gradual entry of foreign banks into the domestic market after the opening of my country's financial market, and the increasingly diversified demands of customers for financial services and products, my country's financial industry is facing a new round of challenges and development opportunities. The financial industry is increasingly paying attention to drawing user portraits and conducting data analysis . Whether it is possible to integrate and analyze the ever-growing structured and unstructured data sources, release the value of data for user profiling, and achieve banking business innovation has become a key factor in determining the future success or failure of banks. The financial industry has achieved data-driven development, accurately outlined user portraits, and built an efficient financial customer analysis system from 0 to 1. This is a whole process from demand analysis, event indicator design, data access, to multi-dimensional analysis.

On the data-driven path, Formax Group has achieved:

• Build an efficient financial customer analysis system to achieve data-driven product optimization and scientific decision-making;
• Accurately outline customer portraits and enhance risk control models through multi-dimensional monitoring;
• Build a refined operation system with a virtuous cycle, accurately track channels , and observe and measure marketing effects in real time.

The following will introduce in detail how Formax Group accurately outlines user portraits and builds an efficient financial customer analysis system from 0 to 1, including the entire process of Formax Group's demand analysis, event indicator design, data access, and multi-dimensional analysis. (Note: Due to suspected commercial secrets, the data involved in this article are all virtual).

1. Sort out the needs

Scientific data collection methods should be based on the business needs of the enterprise. Formax Group divides its business lines into divisions. Each division is responsible for a business line and has independent operations, products and technical personnel. The business department hopes to connect behavioral data with business data to achieve more refined operations.

Formax Group APP sorts out its main data analysis needs based on its actual situation:

1. Demand for detailed analysis of the user situation of the public platform (i.e. the Financial Circle APP under Formax Group)

The needs of all business lines of Formax Group are coordinated by the group as a whole. Through data analysis, we hope to evaluate the usage of the public functions of the entire platform, including:

• User situation: including unique visitors, page visitors, number of active users, number of new users, number of registered users, number of active users of each business, etc.;
• Product usage: including average usage time, visit time distribution, average number of pages visited per person, bounce rate, PV of main pages, etc.;
• Core function conversion funnel: including registration process, card binding process, deposit and withdrawal process;
• Private message usage: including the number of different types of people, the number of unsuccessful uses, the distribution of private messages, the distribution of private message sending time, etc.;
• Use of homepage function modules: clicks on various modules and functions, etc.;
• User attention situation: The types of attention are divided into favorite stocks, experts, column subscription accounts , and the number of followers of different types, the number of new followers, etc.

2. Refined analysis requirements of Life platform

The Life platform is a platform within the Formax Group that has some e-commerce characteristics. Users can use points or currency to exchange for goods. The data requirements for the Life platform include:

• User and product usage: including PV, UV, number of new users, duration of use of the function, duration of stay on different pages, bounce rate, etc.;
• Product transactions: including browsing status of each category and each product, transaction status of each category and each product, transaction conversion funnel, product repurchase status, etc.;
• Order data: including order quantity, average shipping time, average delivery time, etc.

3. Demand for refined analysis of foreign exchange business lines

Forex is one of the channels on the Formax Group APP. In addition to providing basic foreign exchange trading functions, it also provides Copymaster (Formax Group's foreign exchange copy trading community is Formax Group's first socialized product for global financial transactions, bringing together global foreign exchange investment experts.) Foreign exchange trading tools are screened and ranked based on real trading returns on the platform. Ordinary investors can use the "copy" function on specific foreign exchange investment experts in the hope of achieving maximum returns. Therefore, there are two types of roles in this function, one is a high investor and the other is an ordinary investment user. In addition to some basic product usage, they are more concerned about the analysis of two types of roles accompanying transactions.

Enterprises are very concerned about the product usage of two types of roles and hope to understand:

• Product usage: PV, UV, new users, number of registered users, number of account opening users, number of depositing users, page views, clicks on core functions, usage time, etc.;
• Talent investors: number of people who applied for talents, number of new talents, number of people who gave up talents, number of people who are tying talents, talent profit ratio, and core behavior statistics of talents;
• Ordinary investors: number of copies, number of successful copies, copy amount, etc.:
• Transaction data: number of depositors and withdrawors, deposit and withdrawal amounts, number of traders, transaction amounts, and trading products
• Core funnel: registration and account opening process, deposit process, transaction process, copy process

4. Demand for refined analysis of P2P financial management business lines

In this regard, we hope to pay attention to:

• Product usage: PV, UV, number of new users, clicks on core functions, duration of stay on different pages, bounce rate, etc.;
• Transaction status: deposit and withdrawal status, number of traders, transaction amount, reinvestment status, coupon usage, subsequent behavior after investment maturity, bond transfer status, etc.;
• Core funnel situations: registration conversion, financial management purchase conversion, bond purchase conversion, transfer application funnel, etc.

5. Demand for refined analysis of stock business lines

Similar to foreign exchange, using the Forbag ​​stock portfolio tool, you can buy or sell portfolios created by professional investment managers or folk experts with one click.

• Push tracking: receive push , click push link
• Community information: click, load, comment, like, share, distribute
• Trading: deposit and withdrawal, stock buying, selling, order cancellation, and short selling
• Copy trading behavior: the experts who copy orders
• Core conversion funnel: account opening process, deposit process, withdrawal process, etc.

2. Event indicator design

Based on the above requirements, Sensors Data made recommendations on event design solutions based on its actual business situation and actual data analysis needs.

Refined analysis of user needs for public platforms (i.e. APPs under Formax Group); event design includes startup and exit, APP browsing page, APP element click, APP activation, registration & login, real-name authentication, bank card binding, deposit & withdrawal, sharing, etc.

• In response to the refined analysis needs of the Life platform, we designed events such as browsing pages, submitting orders, successfully paying for orders, shipping and receiving goods, etc.
• Regarding the refined analysis needs of the foreign exchange business line, we design events including the foreign exchange account opening process, application for foreign exchange high investor events & giving up high investor qualifications, purchase of foreign exchange products, purchase of foreign exchange guaranteed income products, and foreign exchange copy trading.
• In response to the refined analysis needs of the P2P financial management business line, events such as clicking on financial products, submitting investment information, paying for investment projects, successful investments, investment expiration, receiving coupons, and debt transfer are designed.
• According to the refined analysis needs of the stock business line, we design events for high-investor copycats for each step of account opening, browsing stock information, liking, commenting, sharing, and publishing of information, placing orders, canceling orders, and completing transactions.

The above events contain rich attributes, which are combined with user attributes to mark the behavior and user characteristics when the event occurs. For example, in foreign exchange copy trading events, attributes such as high investor type and high investor ID are included to analyze the copy trading situations of different experts. For example, P2P financial management related events include attributes such as financial product type, financial product name, income method, investment period, annualized rate of return, investment amount, coupon ID, coupon type, coupon amount, actual payment amount, investment income, payment method, etc., so that we can conduct multi-dimensional analysis of investment behavior, understand the investment situation of different product types and different products, the investment distribution of different investment periods and rates of return, and measure the effectiveness of coupon issuance in combination with the behavior of receiving coupons.

Through element click and page browsing events, you can collect all element clicks and page browsing in the APP, and distinguish which element/page the user clicked/browsed through attributes such as the element content and the name of the page it is on. These events, as a supplement to custom events, meet some common needs such as PV, UV, average usage time, average visit depth, bounce rate, click status of each function, etc.

The above event is the preliminary event design plan of Formax Group using Sensors Data. With the development of the company's business, understanding of event design and changes in demand, the event design has been optimized and adjusted. In order to protect customer privacy, only a general idea and framework are listed.

3. Data Access Phase

The data access stage is divided into two parts, namely the access method and the embedding method.

1. Data access method

In this project, in order to ensure the comprehensiveness and accuracy of data access, the data access methods include front-end data collection and back-end data collection:

1. Ordinary behavioral data is collected from the front end;
2. Back-end data collection.

Important events such as deposits and withdrawals, investments in financial products, and purchases of foreign exchange products are collected from the back end. Events such as issuing coupons that are only recorded on the back end are collected from the back end. In some events, such as order submission, some attributes are collected by the front end, such as operating system and geographic information, and some attributes need to be collected from the back end, such as product brand, product category, product price, etc. At this time, the front end transmits the required attributes that can be collected by the front end to the back end, and splices them with the information collected by the back end, and sends them uniformly by the back end.

2. Data tracking specifications

Since the enterprise has many business lines, after determining the access method for each event, requirements are set for tracking point specifications, including:

(1) Standardization of event and attribute names

For each event and each attribute, the English name of the embedding point is defined to ensure the consistency of information transmitted by each business line and each end. In order to distinguish different business lines, additional prefixes are added before the events, such as the prefix P2P for financial management events, the prefix forex for foreign exchange events, and the prefix stock for stock events. This makes it easier to manage events and to use them for subsequent analysis.

(2) Specification of event collection timing

​Clearly define the collection timing for each event. For example, an element click event is triggered when the element is clicked on the front end. Successful transaction events, such as stock trading and successful purchase of P2P financial products, are triggered only after the server returns a success message. Clear collection timing makes it clearer for developers, reduces unnecessary communication, and ensures data accuracy.

(3) Specification of attribute collection scope

1) When the collection attributes of the same event are inconsistent across multiple terminals, it should be clarified

Take the browse page as an example. The enterprise platform has a standard page ID system applicable to all terminals, hoping to conduct a unified analysis of the browsing page behavior of each terminal. Therefore, the web uses PageView and the app uses manual tracking. However, there are many preset collection attributes in PageView that are not available in the app. Such attributes need to be clearly explained to avoid causing trouble to development;

2) The value range of special attributes must be determined

Some events are designed for specific scenarios, and the attribute values ​​​​in them can be exhaustively enumerated. They are also the points that need to be analyzed later, so they need to be clearly listed. For example, if a product manager needs to know the functions of certain specific pages, but does not need to collect information about other pages, then it is necessary to clarify which functions of those pages need to be collected.

In addition, in response to enterprises’ security concerns, Sensors Data provides a private deployment solution to meet enterprises’ urgent needs and protect the platform’s core assets.

IV. Application Scenarios

Scenario 1: Combine with the work order system to restore real user operations and efficiently resolve customer demands

For the financial industry, it is crucial to ensure the security and stability of every user's funds. In financial enterprises, any behavior related to money, such as recharge and withdrawal, will affect the user experience and company reputation once problems arise, and will have a serious negative impact on the development of the enterprise.

Formax Group uses a work order system for customer service, including customer support and help services, processing and resolving customer service requests, etc. After the work order is delivered to the target service desk, the main processing flow includes: responding to customer requests - listening to customer feedback - providing feedback to technical personnel - technical personnel inquiring about the situation.

However, throughout the entire process, the source of all judgments and arrangements made by customer service staff is the customer's description: What operations did the user perform? What abnormal situations occurred? wait. However, from responding to customer requests to processing requests, relying solely on customer statements will result in inaccurate information for various reasons, extend the customer service cycle, and easily cause customer dissatisfaction.

Figure 1 User behavior sequence

Now Formax Group is able to restore real user operations. For example, the personal behavior sequence can very conveniently view the user's specific operation behavior. In addition, the specific attributes of each behavior event are also displayed, such as the callback result of each interface, the reason for failure, etc. In this way, customer service staff can quickly discover problems, give customers reasonable explanations at the first time, and quickly solve customer problems. Achieve the following effects:

1. Visualize user behavior operations to avoid delaying customer service operation cycles due to ambiguous or incorrect user descriptions;
2. Locate abnormal situations in a timely manner to improve customer experience and corporate credibility.

Scenario 2: Data-driven positioning of the best opening screen theme

Formax Group collects the ID of each page and all button IDs and button names on the page and other related attributes to understand every operation behavior of users on the APP. The opening page of the Formax Group APP will display some marketing information or event information to financial users. In the early stages of the APP’s operation, product managers believed that users’ demand for “fund security” was far greater than user experience . Therefore, content such as "professional qualifications" and "multi-country licenses" will be displayed on the opening event page to convey a sense of brand security.

Figure 2 Conversion rate of the opening page for the theme of "fund security"

Through data analysis, it was found that the conversion rate of users from March 15 to March 30 was 2.29%, which was not an ideal result.

Figure 3 User conversion after launching a high-yield page activity page

Later, we tried to launch an activity page with the theme of "high returns", and through funnel analysis, we found that the conversion rate of the Kaiping page targeting high returns would be higher, up to 14.05%. Therefore, designing products based on data rather than people's subjective judgment is a relatively successful revision.

In addition, relying on rich user behavior, in the optimization of other product details, we also abandoned the solution of subjective judgment of personnel, and relied on data to speak. By designing two or more solutions, we selected the better solution through analysis models such as event analysis and funnel analysis.

Scenario 3: Create a full-process refined operation system with user segmentation, precise push, and effect feedback

The high cost of acquiring customers, increasing customer stickiness and extending the customer's lifetime value are the most concerned issues for Internet financial companies, and Formax Group is no exception. It is a very common marketing method to push content to users efficiently and conveniently to awaken dormant customers. The process usually includes the following:

Set the theme and goal of the event - locate the target audience for marketing - reach the target audience with marketing content - observe whether the effect of the push has achieved the goal. The following is a marketing campaign conducted by Formax Group through the Sensors analytics platform.

(1) The first step is to screen out “high-intent users”

In the "User Segmentation" function page of the User Analysis module, locate the target group by conditional filtering. For example, in order to "awaken" users who registered in January 2017 and browsed the credit report page (analysis found that users have a higher retention rate after browsing the credit report page), but did not invest, in order to lock in the target population, you can perform the following operations on the "User Segmentation" function page of the User Analysis module.

Figure 4: Filter marketing target groups on the "User Grouping" function page

(2) The second step is to push information to the “high-intent customer” user group

This group of people can be screened out through the user grouping function, and they can be notified via text messages or pop-up windows on the site, and information can be pushed to the group to stimulate their investment.

Figure 5 After users are divided into groups, push notifications are implemented within the site

(3) The third step is to evaluate the push effect

After completing the information push, operators can conduct multi-dimensional analysis and display the effects of the push in real time. After the Internet finance customer completes the precise push, the user can view the user conversion status again in the investment process conversion funnel to evaluate the push or product optimization effect.

Figure 6: Comparison of the overall conversion rate between the people who were pushed and the people who were not pushed

As can be seen from the figure, after completing the precise push to "high-intent customers", the overall conversion rate reached 24.69%, while the conversion rate of people who did not receive the push was 16.34%, indicating that this was a relatively successful precision marketing.

In this way, Formax Group has built an efficient, convenient and precise marketing platform. On the visual interface, enterprise operations personnel can complete the entire process of refined operations, including multi-dimensional indicator user behavior analysis , user grouping, accurate information push to target groups, and real-time viewing of push effects.