A scientific and complete data indicator system is an important support for enterprises to carry out digital operation management and build a data-driven organization. Through multi-dimensional data indicators, operation personnel can clearly understand the business status, product/R&D personnel can efficiently locate system problems, and management personnel can make more accurate analysis and decisions.
So, how to build an effective and easy-to-use data indicator system on the basis of fully understanding business needs? This article reviews the third episode of Getui TechDay's "Number Governing Training Camp" series of live broadcast courses, and shares "Data Index System Design and Development Actual Combat" with you .
course review
Data indicators and indicator system
A data indicator is a specific type of metadata information. It is a quantified measurement value after subdividing business units, and it is also the intersection of business and data. Data indicators make business goals descriptive, measurable, and disassembled, and can provide quantifiable support for daily development iterations of product operations and guide scientific decision-making.
Data indicators are generally divided into two categories : result indicators and process indicators . Result-type indicators are used to measure the results produced by users after a certain action, and to measure whether the needs of users in a certain scene are met. This result is usually known after a delay, and it is difficult for people to intervene. Process indicators refer to the indicators generated by the user in the process of doing a certain action. The process indicators pay more attention to why the user's needs are met or not met. People can affect the process indicators through specific strategies, thereby affecting the final result. the result of. For example, as far as an e-commerce promotion is concerned, the final sales volume is a result-based indicator, and data such as product page exposure, clicks, and additional purchases are all process indicators. E-commerce operators use operating strategies to increase exposure, clicks, etc. Rate, additional purchase conversion rate and other process indicators will affect the final result indicators.
By analyzing the data indicators on each link of the sales conversion link, business personnel can clearly grasp the business situation
A single data indicator cannot fully reflect the business operation situation. We also need to start from the overall situation and systematically organize scattered, single-point, and interrelated indicators to build a data indicator system . The process of establishing the data indicator system is actually a process of thinking about the nature of our business. A set of scientific and complete data indicator system can measure the quality of business development, help us see the overall business situation through a single point, and solve business problems at a single point through the overall situation.
Data indicator design and development
To build a data indicator system, an enterprise must first sort out the corresponding data indicators based on business goals. We recommend referring to the OSM model to disassemble business goals and complete the design of data indicators.
OSM model
Among them, "O" refers to the target Objective, "S" refers to the strategy Strategy, and "M" refers to the measurement indicator Measurement.
Taking the e-commerce operation scenario as an example, the operation goal (O) of the e-commerce platform is often to increase GMV. According to the formula "GMV = number of paying users x unit price per transaction x user purchase frequency", then its promotion strategy (S) and corresponding metrics (M) may be:
✦ Increase the number of payment users
The strategy (S) is to give new registered users 9.9 limited-time special benefits, and the metric (M) is the number of new registered users.
✦ Increase the unit price per transaction
The strategy (S) is to sell assorted items, and the metric (M) is the average unit price per order.
✦ Increase user purchase frequency
The strategy (S) is coupon marketing on holidays, and the metric (M) is the frequency of orders placed by users.
The OSM model disassembles grand and abstract goals into a series of specific, implementable, and measurable behaviors, and is applicable to many scenarios such as product operations, user operations, performance management, and business operations.
Taking the user analysis scenario as an example, through the combined use of the OSM model and the UJM user journey map model (User-journey-map) , operators can quantify the entire process experience of users from clicking, browsing to adding purchases, placing orders, and sharing Management, find out the key links that affect the final purchase conversion rate of users, and optimize them accordingly.
For another example, in the product operation scenario, product personnel combine the OSM model with the HEART model to quantify and evaluate users' experience with specific product functions through a series of indicators, and provide data support for product iterative upgrades.
Index classification
The design of the data indicator system is a relatively complicated task, and it also requires enterprises to carry out top-down indicator classification according to their own strategic goals, organization and business processes , so that different roles such as management and business personnel can understand the data more efficiently. The meaning of data indicators, and quickly locate related problems through the fluctuation of data indicators.
Generally speaking, we divide data indicators into three levels: the first key indicator (also known as "North Star Indicator"), the first-level indicator and the second-level indicator . For example, the first key indicator of taxi-hailing apps is the order completion rate. By dismantling the order completion rate, two first-level indicators can be obtained, namely, the number of orders issued and the number of orders completed; Secondary indicators such as the number of canceled orders and the number of canceled orders by passengers. For customer service personnel, it is more necessary to pay attention to secondary indicators, follow up to understand the reasons why drivers and passengers cancel orders, and solve user experience problems for drivers and passengers. As for the operators of taxi-hailing apps, they need to pay more attention to first-level indicators such as the number of orders issued and the number of completed orders, and improve the corresponding indicators through operation and incentive measures.
Indicator design
Before designing an indicator, we need to understand several major components of the indicator: dimension, measurement, statistical cycle, filter conditions , etc. Dimensions are descriptive data, which refers to the statistical environment of indicators such as regions, product names, and product types; metrics are numerical data, such as product sales, account balances, etc.; statistical periods refer to the time range for calculating indicators, such as This month, this quarter, this year, etc.; filter conditions refer to the conditional restrictions for calculating indicators, such as valid status, non-working days, etc.
The constituent elements of the indicator determine the production logic of the indicator. According to the different components and production logic, data indicators can be divided into atomic indicators, derived indicators, composite indicators and other types. Among them, atomic indicators refer to the measurement of a certain business behavior event, such as the number of transactions, transaction amount, number of transaction users, and account balance; derived indicators refer to the derivation of dimensions, statistical periods, or filter conditions based on atomic indicators , such as the amount of account consumption in the past week, the account balance of the previous year, etc.; and the composite index is more complicated, generally calculated by adding, subtracting, multiplying and dividing multiple indicators, such as the monthly average GMV in 2022, the annualized investment income etc.
Metric Metadata
According to the different types of indicators and production logic, enterprises can clearly sort out the data sources required for production indicators and what kind of data model needs to be built to calculate the indicator results. In order to manage the indicator life cycle more standardizedly, we suggest that enterprises can output an indicator metadata specification, which clearly lists the indicator name, indicator code, indicator catalog, indicator classification, business caliber, technical caliber, indicator responsible person, and indicator data. Important content such as update frequency and description information provide more detailed guidance and reference for data developers, indicator users, and indicator maintainers.
Metric Review and Development
After completing the design of the indicator model, indicator content, etc., the data analyst/data warehouse architect will hold an indicator review meeting to fully discuss and reach an agreement with the data development/business personnel on the definition of indicators, business caliber, technical caliber, update cycle, etc. Opinion.
Business personnel are the demanders and users of data indicators, and can put forward constructive opinions on issues such as what are the derived indicator dimensions, what is the statistical cycle, and what indicators are processed from composite indicators; data developers have a better understanding of the data sources of the enterprise According to the current situation, it is possible to give professional advice on technical issues such as which data models in the data warehouse for the derived indicators are processed and output.
Based on the feedback from multiple parties at the indicator review meeting, the data analyst/data warehouse architect responsible for indicator development can optimize and iterate indicator metadata and indicator production logic, and officially start indicator development.
Experience summary
The data indicator system is a very important data asset of an enterprise. Combining our experience accumulated in the process of data governance and indicator system construction, we recommend that you grasp the following three key points in the process of building a data indicator system:
①Follow a set of standard and standardized indicator construction methodology, and design an enterprise-level data indicator system;
② There is a unified process control mechanism to comprehensively control and manage the life cycle of data indicators;
③Build a unified indicator management platform, conduct centralized management of data indicators, and accumulate indicator assets.
Featured Q&A
During the live broadcast, everyone exchanged views on the course content. This article selected the wonderful questions from the live broadcast room for Q&A sorting out.
Q1: How to use unstructured data and semi-structured data to build an indicator system?
This requires specific analysis of specific issues. Generally speaking, when we establish an indicator system, we aim at structured data. For unstructured data, we also need to first govern and convert it into structured data, and then proceed to the construction of the indicator system.
For example, for data in video format, we need to use video recognition algorithms to convert video format data into structured data, incorporate it into the entire data warehouse system, and then build an indicator system in a targeted manner.
Q2: How to measure the quality of the indicator system?
A high-quality indicator system can not only clearly reflect the current status of the company's operations, but also be used by people at different levels to help companies/organizations develop better.
At present, many companies have built data dashboards and data cockpits, which combine and display various data indicators according to different themes, helping management, middle-level leaders, and business personnel to view the business status of the company, analyze and judge business problems, and measure performance. and achievement of business objectives.
In order to help enterprises analyze and use data more conveniently and efficiently, DIOS, a daily data management platform, adopts a low-code design concept, so that business personnel in different departments such as marketing, HR, and finance can also flexibly create data dashboards and data portals And other data applications, bring out the value of the data indicator system, and respond to the diverse data needs encountered in business scenarios in a more timely and rapid manner.
Q3: Can metrics be built using unstructured data in the data lake?
OK.
The data lake has changed the way that the data warehouse first processes the data and then uses the data. The data lake emphasizes storing the data first, and then considering the specific data processing method when the data is wanted to be used later.
A large amount of semi-structured and unstructured data, such as images, voice, and video, are stored in the data lake. How to efficiently extract valuable information from unstructured data is the difficulty for us to use unstructured data in the data lake to establish an indicator system.
Pay attention to the WeChat public account of Getui Technology Practice,
The background replies with "indicators" ,
Get the live courseware of this issue~
Next issue preview
The fourth phase of the TechDay Numeracy Training Camp is coming soon!
On November 30th (next Wednesday) from 19:30 to 20:30 in the evening , the senior data product manager from Gezhishu Platform Department will sort out the methodology and core strategy of the company's labeling system construction in detail , and share the benefits of labeling data in depth. Capitalized operation and management experience, interpretation of the scene application cases of the portrait label system.
