The idea of building data kanban

Who | 0 x00 kanban with data

Before we talk about a subject, we should first ask ourselves two questions:

  • One, what does this thing do?

  • Second, who is it for?

For example, statements, people think, it’s natural to think, this is for executives or investors to see.

However, duan Yongping, known as “China’s Buffett”, said when talking about his investment experience, he would say that he almost never read statements, and if he really wanted to know, he would ask professionals to read them and give him a conclusion. It’s more like a test sheet of diagnostic indicators that doctors can use as a reference, but it’s not a diagnosis.

Therefore, the function of data kanban should be a table or chart used by professionals to present the company’s business data, so as to facilitate the relevant personnel within the company to grasp the business development situation. The users of data Kanban should also be mainly business leaders and front-line operation employees, assisting the intelligence needs of senior executives.

With layered user groups, we can draw a clear portrait of the character:

| 0 x01 data report design

The core content of data Kanban is the data report, and other functions, such as data interpretation, index insight, are an extension of the report.

But the data report is not a hot brain, draw a sketch can start dry, especially for complex business, only in the standard analysis system, how to set up screening conditions, view statistics, select the appropriate chart, to build fast, query speed is not slow. Some relatively basic and important reports, usually also need to carry out the corresponding data test, to ensure that the content presented is correct.

As the saying goes, “Make the right decision based on the right data.” If the data is not correct, then what analysis?

In general, there are three key steps in the design of data reports:

  • Design analysis system;

  • Define data indicators;

  • Adjust the data report.

【 Design analysis system 】

Business is the core of data report design. If you want to have a clear analysis system, the key point is to get through the link of data analysis.

For example, in order to analyze the growth of users, AARRR model is used to obtain the corresponding data of user activation, retention, source and so on, and statistics the transformation of each stage. To analyze sales profit, you need to know data such as profit margin, inventory rate, product category, quotation strategy, and market fluctuation.

Normal business will have a mature analysis system, which can be more with the product operation market students to talk.

[Defining data indicators]

In most cases, the data department will have its own metrics library, but when it comes to building kanban, there are three main concerns: caliber, dimension, and statistical period.

Unclear data caliber is probably the most common situation encountered by business people and the most contentious with the technical team. Therefore, for some relatively common indicators, such as UV, GMV, ARPU, etc., it is necessary to have a clear definition, as much as possible with the business consensus, and in the internal document system can be retrieved at any time.

After defining the data caliber, you need to observe the main indicators of the current service and gradually separate the observation dimensions, commonly known as “drilling down”. More normal metrics, providing the ability to drill down, are standard, because deeper dimensional analysis of the data is possible to get accurate answers. However, not all dimensions should be included in the report, and some business-irrelevant dimensions do not need to be displayed.

Finally, there is the statistical time cycle, which is two frequencies by day/hour under normal circumstances. If the business changes quickly, then 15 minutes or even real-time data need to be seen. But generally speaking, the faster the frequency of updates, the higher the technical requirements, this still depends on the team’s current technical capabilities.

[Adjust data report]

Data reports are different from products. In order to provide the most valuable information per unit area, the MVP principle should be strictly followed and the product should be iterated with the minimum set of features.

But to have a certain psychological expectations, because very few statements are designed, no longer change. In the early stage of the launch, frequent adjustment of indicators, statistical range, screening conditions and permission control are very common problems. Therefore, it can be pushed to the relevant business side through Excel in the early stage, and the problem of going online can be considered when the business is stable.

Common questions to consider when reporting online include:

  • Default query time, whether there is a null value at the beginning of the month;

  • Data sort rules, default according to which column sort;

  • Setting of screening conditions, which conditions are most suitable for users to open;

  • Check the statistical indicators, whether there is a need to merge, de-weight situation;

  • Handling of outliers, such as null values, limits, etc.

| 0 x02 kanban design data

When we can independently build data reports, how to efficiently and quickly build a report with good interactive experience through the visualization capabilities provided by the tools, the test is the accumulation of past experience.

For example, I recommend the following report design principles:

  • Show the core metrics, prioritizing speed over just cool results;

  • The data information between different blocks should have correlation, which can reflect your thinking of analyzing the problem;

  • Content should meet business requirements, do not display irrelevant information;

  • The change trend of indicators can be reflected in charts, and the key points should be explained in words.

  • Comparison of different data indicators is the most intuitive way to find problems.

As can be seen from the above information, it is important to choose the right chart. Here is a chart selection suggestion that is common on the Internet:

But there are a few other design tips worth learning:

  • Keep your table organized: field alignment, date formatting, block scaling, and proper height; And so on;

  • Make the table more intuitive: The background color of the cells should be compared to the content, so using zebra stripes can speed up browsing. Avoid dark, bold borders, which distract from the frame. Highlight the key data; And so on;

  • Use simple and clear hints: Hover hints, default action hints, pass-through copywriting, etc.

  • Folding detailed data: Although detailed data is the core content of the report, if there is too much data, the content to be expressed in the report is not intuitive. Therefore, some information needs to be hidden. For example, in the tree table, only the first column is expanded by default. For example, skip to the details page drilled down with more options; For example, hide more exploratory metrics through the field filtering capabilities of reporting tools.

The next step is to make specific optimizations for different users. Because we may have dozens of different business processes and hundreds of relevant data indicators, but not everyone needs to see them, which data indicators and measurement methods to choose need to be determined through communication with the business team. For example, some indicators are the appeal of core business (DAU, etc.), and the former is always the core of the development of the company compared with some indicators that are more oriented to the management level, and the demand priority is higher than the latter.

The first is the report design for senior executives. The design principle is simple: “concise, results-oriented, prioritized”. The design of such reports tends to display pure digital content rather than charts, because the executives need to grasp the current situation of the business in the first place, and charts increase the cost of acquiring key information. The other is that executives are so sensitive to business data that they don’t need curves to show patterns, they basically just need numbers to quickly determine the health of the business.

Senior management is more sensitive to data than front-line development, so don’t try to muddle through.

The second is the report design for the business, highlighting a “vertical segmentation” feature. Because there are many factors that affect business, business personnel are more “focused on process” than executives are “focused on results”, so they are used to using ICONS such as curves and funnels to show the change trend and change rule of data. So what the business is asking for is basically 360 degrees of drilling and correlation, the degree to which you can get to the bottom of the data.

Designing reports for business is the most time-consuming, such as visual features, block linkage, dimension driller, exception highlighting, annotation, mobile adaptation, etc. There are already many problems to be considered outside the data analysis system.

Welcome to click here to read further: Towards the Highest level of data: Data visualization experience in detail.

| 0 XFF data push

After the report design is completed, in order to save time, remind risks, daily routine, etc., the report needs to be configured in the form of pushing.

There are three ways to push: email push, chat tool push alone, chat tool group push.

Email push is a common push method, and the configuration is flexible, so that business personnel can check it immediately after coming to the company every day.

Chat tool push, now nail nail and other enterprise communication software basically support, configure a push port. This approach has the advantage of being more efficient and avoiding information being overwritten by redundant messages.

Sometimes, if data is abnormal, corresponding information push can be configured to remind developers or business personnel to check.

Sometimes, report push is a necessary thing, because when the information is very complicated, it is easy to miss something. In this case, the simple push function represents part of the data-driven capability. The way to increase business value is not just algorithms or analysis reports, but human experience plus engineering.

Want to see more?

Click below for more dry goods ****👇

= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =

Recommended reading:

Kuaishou’s data governance practices start with model specifications

Data warehouse architecture and multidimensional data model design

Data index system construction method

Common modeling methods and examples in the field of data warehouse

About building and optimizing data warehouse architecture and model design