foreword
The group standard " Software R&D Efficiency Measurement Specification " initiated by Zhongguancun Zhilian Software Service Industry Quality Innovation Alliance and China Software Association Process Improvement Branch has been released at the TiD 2022 Quality Competitiveness Conference.
The "Specification" expert group is composed of performance experts and well-known consultants from companies such as Tencent, JD.com, Simayi, Hengqin Life Insurance, and CETC. Forty companies including Huawei, Tencent Cloud, Baidu, JD.com, NetEase, Smayi, Ping An Bank, Everbright Bank, ZTE, and Xinhua III participated in the co-creation.
"Software R&D Efficiency Measurement Specification" has established a R&D efficiency measurement framework and index set suitable for the domestic software industry according to local conditions, providing a basis for the R&D management and improvement of the software R&D team, and at the same time providing a basis for the construction of software R&D data platforms, data exchange and communication in the industry Collaboration provides the theoretical basis.
Ren Jinglei, a member of the expert group and the founder and CEO of Smarti, shared the interpretation of the key points of the "Software R&D Efficiency Measurement Specification" at the TiD 2022 conference, and discussed the open source implementation of relevant R&D efficiency indicators based on cases.
Against the backdrop of a contracting economic environment, the topic of R&D effectiveness has attracted more attention. In the middle of this year, Meituan's internal meeting summarized three key propositions for 2022, including systematic cost reduction and efficiency increase, and seeking benefits from leaner management .
As the R&D organization that is the cost center of a large number of technology companies, there is no need to repeat the attention and expectations that its R&D efficiency has received. But what posture is used to practice R&D efficiency is scientific and effective? How to implement the most controversial measure of R&D effectiveness?
Regarding this issue, there have been many exchanges and discussions in the industry in the past few years. The "Software R&D Effectiveness Measurement Specification" released this time is also based on the real practice of 40 companies and the thinking experience of the members of the expert group.
Through a series of tasks such as sorting out measurement definitions, formulating measurement frameworks, clarifying measurement principles, and proposing measurement methods , the "Software R&D Efficiency Measurement Specification" hopes to accumulate professional knowledge in the field of R&D efficiency and help the R&D team to achieve a reading of more than 10,000 "volumes"——Improve Cognitive, avoid using wrong postures such as introversion, and roll out R&D performance anti-patterns.
01 Index System of "Software R&D Efficiency Measurement Specification"
The following picture is the E3CI software R&D efficiency framework defined in the Specification .
We can see that R&D effectiveness (the E in E3CI) is divided into three dimensions, which are the efficiency from an engineering perspective , the effect from a business perspective , and the excellent ability to continuously achieve efficiency and effectiveness .
The R&D effectiveness is achieved by measuring and obtaining cognition (C), plus practice improvement (I).
The knowledge gained is measured in five cognitive domains: value delivered, velocity, quality, cost, and capability . "Software R&D Efficiency Metric Specifications" provides a structured review of common metrics based on cognitive domains and software R&D links.
However, the indicators themselves are static and waiting to be called, and the listing of the indicators themselves does not allow the R&D team to understand how to use them. Therefore, the "Software R&D Efficiency Measurement Specification" also provides an index model for software R&D efficiency measurement to guide the smooth implementation of R&D efficiency measurement.
The following indicator model diagram can be interpreted from two aspects: demand sorting and tool realization .
1.1 Demand sorting
From the top value stream to the measurement indicators, it is the requirement sorting process in the measurement design phase.
Metrics exist to satisfy certain information needs in the business value stream, the absence of which prevents us from making credible analyzes and decisions.
What is emphasized in the demand combing link is to clearly define these information needs in combination with the actual situation . Depending on the business nature, stage, strategic focus, and other attributes of the team, the information needs will naturally vary. If you start measuring without understanding why you want to do it, if you want to be comprehensive, or copy other people's successful cases mechanically, it will inevitably make the measurement become tasteless.
In the previous exchange " Understanding the Essence of the GQM Method for Measuring R&D Effectiveness " , Mr. Ru Bingsheng also mentioned that many companies measure for the sake of measurement, and start collecting data directly from the indicators as soon as they come up, rather than from their own specific and clear Starting from the goal , this is a common pitfall in the measurement of R&D effectiveness.
Although it cannot completely repeat the previous route, some industry methodologies can serve as a compass to help R&D teams define their own measurement needs comprehensively and clearly. The GQM method we introduced before is to drive the R&D team to dismantle layers through the three progressive levels of goal-problem-indicator , and think about what kind of measurement index system can truly meet the needs.
1.2 Tool implementation
From the bottom data source to the measurement index, it is the process of tool realization . Data sources include various R&D tools and manually entered data, all of which have accumulated a large amount of R&D data. However, due to the large volume, dispersion and low degree of standardization of the data, it cannot be used directly. Therefore, it is necessary to collect a large amount of R&D data into the data lake for governance and extract credible data sets .
Compared with the requirements combing process mentioned above, the tool implementation process is actually more reusable.
One of the common misunderstandings in the practice of R&D efficiency is to be lazy in the process of sorting out non-standard requirements, and to be too diligent in the process of implementing standardized tools, reinventing the wheel . In fact, R&D teams can use open source tools to build R&D performance measurement tools at low cost.
02 Open source implementation of standard indicators
As mentioned earlier, measurement tools need to achieve data collection and governance. What is it specifically? Here is a simple example of the functions of the open source R&D data platform Apache DevLake .
- In terms of data collection, it is necessary to deal with the concurrency, fault tolerance, scheduling, etc. of each data source, design a variety of data pull or push methods, and design hierarchical storage from raw data to abstract data to meet both speed and reliability requirements.
- Data entities in different tools of the same type (such as JIRA and TAPD) may not be consistent. If multiple tools are used at the same time or tools are switched within the organization, modeling is required to solve the problem of consistency between tools through abstraction, mapping and association .
Abstract data into the domain model of each link in the R&D process
- For data query/reading performance and management convenience, it may also be necessary to aggregate data and store it in a highly correlated warehouse.
- After screening credible data sets, indicators need to be extracted and combined into data dashboards based on scenario requirements to meet the information needs of analysis and decision-making.
Data Kanban for continuous integration scenarios
03 Summary
The so-called R&D effectiveness measurement seems to be just an indicator and a number extracted from the data set. But when viewed in the indicator model, an effective measure of R&D performance is much more than just a single number.
If your team is implementing R&D performance measurement, I hope that the " Software R&D Performance Measurement Specification " can become a reliable tool manual to help your team avoid pitfalls.
At the same time, it is hoped that as the R&D efficiency measurement practice in the industry becomes more mature, the " Software R&D Efficiency Measurement Specification " will also grow synchronously, continue to be enriched, and continue to inject valuable knowledge and thinking into the industry. If you want to know more about R&D effectiveness measurement, please download " Software R&D Effectiveness Measurement Specification "
Welcome readers and friends, click the text below to download!