QCon Global Software Development Conference | Large-scale team R&D efficiency continuous improvement practice

On the morning of October 21st, at the 2021 Qcon Global Software Development Conference (Shanghai Station), ONES CTO Feng Bin delivered a keynote speech titled "Practice for Continuous Improvement of R&D Efficiency of Large Teams", sharing R&D with many industry leaders and practitioners Practical experience of efficiency improvement.

The following is the main content of the speech.

According to statistics, on average, 75% of the reasons for the failure of IT projects in large R&D teams include delays, over budgets, unsatisfactory functions, and poor software quality.

Therefore, if these problems can be effectively solved, the success rate of software development is expected to increase greatly.

Combined with the practice of ONES, we found that large R&D teams face the following challenges:

Challenges faced by large R&D teams

01 Big difference in work effect

Because different teams use different tools, platform-based solutions are difficult to apply uniformly, and team data is scattered, resulting in low team collaboration efficiency and difficulty in improving overall performance. Teams with good grades usually rely on several experts, but "experts" are difficult to replicate to other teams.

At the same time, project progress is a core indicator to measure team efficiency, and requires R&D team collaboration to advance. However, the predictability of task progress for medium and large teams is low.

02 Low business satisfaction

In daily work, there will always be business departments who will question the R&D team: Isn’t this requirement very simple? Why did it take so long?

This reflects the problem of "excessively long demand lead time", that is, the time from the proposal to the delivery of a demand is too long to meet the expectations of business parties or customers.

At the same time, due to various inevitable objective problems such as technical risks and emergencies, the project schedule promise is easily broken, and at the same time, it may lead to a high rework rate, causing the team to complain.

03 Improvement measures are difficult to implement

The implementation of improvement measures requires the cooperation of all team members. However, in the early stages of improvement, managers often encounter such problems: team members think "We are already very busy, and we have to spend extra time and energy to improve"? As a result, the team thought the matter was "high cost of execution."

Not only that, improvement often means "reform" within the team, which may hit some people's original roles or positions, causing team members to question the improvement measures, which leads to high resistance to implementation. The cost of improved training and communication is high, especially for large and medium-sized teams, which requires continuous training, which is costly and slow to take effect; the implementation effect after training is difficult to track. These factors comprehensively affected the implementation of the team's improvement measures.

Based on the challenges faced by large R&D teams, we believe that R&D management improvement must be promoted by "non-reliant means", but by setting standards and implementing them in tools to achieve automated management.

Before implementing improvements, we must first find out what are the underlying factors that affect R&D efficiency.

Factors affecting R&D efficiency

01 Workload and work rhythm

As mentioned earlier, in the process of project execution, the R&D rhythm is often broken due to sudden emergency tasks, resulting in the accumulation of R&D tasks. This means that some of the requirements put forward by the business departments have not been completed by the R&D department. If things go on like this, serious workload accumulation will occur and a situation of out of control will appear.

02 Establishment and implementation of standards

Standards are the most important means to ensure the scale of the team. An effective standard should be a clearly defined work process and a completion standard for each link that is clearly defined. Moreover, managers must confirm that the team is working around the standard.

03 Improved sources of resistance

How to make team members understand and accept improvement?

In fact, before improving specific things, people’s perceptions should first be changed. When the improvement of multiple factors is involved, the resistance will be doubled. It may involve people, things, tools, etc., such as organizational structure adjustment, work habits change, and tool implementation.

In view of the above key factors, let's take a look at how to effectively implement improvements?

Continuously improve the landing practice

The improvement process can use the PDCA cycle theoretical framework.

The PDCA cycle is also known as the Deming cycle. The four letters of PDCA stand for: Plan (plan), Do (action), Check (check), Act (process). Deming is an American quality management master. He believes that high quality does not come from product inspection based on results, but from continuous improvement based on the process. Later, his ideas were not only used in quality management, but also widely used in the field of enterprise management.

PDCA cycle

1 Planning stage

How to change people's perception? The visualization of the project is the first step.

Scientific data visualization can directly give feedback to the team, provide a basis for rational analysis, and provide a visual impression for perceptual cognition.

The following figure is the distribution diagram of the team demand cycle time displayed in the ONES report. This is a highly discrete distribution diagram, indicating that the predictability of the delivery time of the demand is very weak.

Demand cycle time distribution chart

Another "status trend chart" shows that the team’s demand for research and development has been increasing, but the speed of release cannot keep up with the speed of demand for research and development, which may eventually lead to the accumulation of team demand and reduce the efficiency of research and development.

State trend graph

However, continuous visualization requires online and structured work content. Take the management of user stories as an example. Through the ONES R&D management tool, managers can customize the user story fields according to the actual scene of the team for the executives to fill in, so as to ensure that the R&D data is recorded in a structured manner, which is convenient for later data visualization. Effectiveness analysis.

Structured work content

In addition, the planning stage also needs to limit the number of parallel tasks, find the bottleneck in the process; identify different work processes, isolate the organizational structure and human investment, and ensure the efficient use of resources. If you encounter multiple resistance factors during improvement, you should give priority to the details with fewer influencing factors and start to improve. The advantage of this is that you can quickly give the team positive feedback and increase the team’s confidence in the improvement. The premise is that the manager is in the team. There is a systematic understanding of the overall situation to be improved.

For example, through the Kanban tool, you can intuitively understand that the team has very few "tasks to be developed", but the tasks of "development", "to be tested", and "testing" are piled up. At this time, managers need to adjust resources to clear the blockage. .

ONES Project Kanban View

2 Execution stage

First, establish a simple and easy-to-understand standard-the standard is easier to be recognized and implemented.

A simple standard does not need to cover as many scenes as possible. A large and comprehensive standard can easily lead to difficulties in implementing the standard. A simple standard can cover 90% of the scenes. The standard is not to tell everyone what the "all" work is, but to "at least" what to do.

Secondly, use tools to digitize the workflow and complete the standards inline. "Embedding" refers to the implementation of the work and status transfer rules of each role in the team into the tool to reduce the interference of human factors and the deviation caused by noise.

ONES Project workflow engine

Thirdly, the team is encouraged to use tools to carry out daily work, so as to realize that the R&D process is truly implemented in accordance with the standards.

Finally, use tools to automate as much as possible. As shown in the figure, it shows a classic automatic flow of demand and task status linkage. When multiple subtasks associated with the same requirement are completed, the requirement is automatically completed.

ONES Project Automation Engine

From here we can see that the benefits of these practices are: through tools to guide the team to complete work efficiently according to standards.

3 Analysis phase

After the improvement measures are implemented, managers need to continuously monitor efficiency data.

For example, comparing the "demand cycle time distribution map" before and after the improvement, it can be analyzed that the demand cycle time of the team after the improvement is concentrated within 20, which proves that the team's demand delivery has become more predictable.

(Left) Before improvement (Right) After improvement

Another example is the "status trend chart" shown in the figure below. It can be seen that after the improvement, the requirements released by the team gradually increase, and the new requirements continue to be consumed, which proves that the research and development efficiency has been significantly improved.

(Left) Before improvement (Right) After improvement

4 Adjustment stage

The standards are not static, but need to be reviewed regularly to ensure that the standards are better adaptable in the process of team development and changes, and the results of the review are incorporated into the next cycle for planning.

Finally, summarize the above content in two points:

1. To improve efficiency, we must pay attention to the arrangement of workload, whether there are simple and clear work standards, and start with a single factor.

Second, the standardization of tools will help to continuously implement standards and visualization on a large scale, and then achieve a state of continuous improvement.