Case analysis: using "measurement" to improve enterprise R & D efficiency ｜ ones talk

6 month 17 Japan , stay 「 Lean Software Engineering Conference 」 On ,ONES Dongxiaohong, a consultant for R & D efficiency improvement, made a presentation entitled 《 Interpretation of R & D effectiveness measurement scenarios 》 Speech . Through two scenarios of efficiency improvement , Discuss how to pass 「 Measure 」 Help enterprises solve specific problems , Improve R & D Efficiency .

The following is the main content that dongxiaohong shared that day .

What is R & D effectiveness ？

In consultation , Customers often ask ：「 How to improve R & D Efficiency 」？ however , The understanding of R & D effectiveness varies greatly ： Some people think that effectiveness is a set of quantitative indicators , Some people think that efficiency is the saturation of human resources , Others believe that efficiency is engineering practice and so on .

Peter Drucker, the master of management, once said ：「 Efficiency is doing things the right way , And efficiency is doing the right thing .」 We understand R & D effectiveness , Is dynamic 、 Do the right thing sustainably . The so-called doing the right thing , It means that we find the goal individually or as a team 、 Direction , Can bring value to customers , So as to help our organization realize business value . meanwhile , Complete the work according to the standard process , Do more 、 Fast 、 Good again 、 Another province , This is what we understand 「 R & D effectiveness 」.

When we talk about R & D metrics , We are used to talking about it 「 It's not something 」—— It's not just a set of data , It is the information behind our data ; It is not a pre-existing indicator , It is to determine the problem to be solved and the goal to be achieved , To design a set of indicators ; It is not a fixed activity of the whole software development , It is continuous feedback in lean thinking 、 The idea and idea of continuous improvement .

Enterprise efficiency improvement case sharing

Next , Share two scenarios of the same company , We hope that the customers can understand how we passed 「 Measure 」 To solve specific problems , And what good practices we have while solving problems . The selected case is an Internet enterprise , The company size 2000 people ,9 Business units , Technical personnel 900+.

Scene one

The pain point of this scenario is stability . The company structure is 「 Small front desk + Tai Chung Tai 」, The capital loss caused by the failure is huge , CIRC interviewed for rectification , Partner complaints , Lost a lot of channels . For the R & D side , They don't have the data 、 Tired of fighting a fire , The stability quality of the current system cannot be clearly stated , Always carry the pot for a third party . This was the pain point of the company at that time .

Based on this sore point , We have set the following goals ：

• Each business unit shall establish stability baseline data —— That is to say 「 Where are we 」？

• Each business department sets stability goals according to its own business —— That is to say 「 Where are we going 」？

• Repeat the disk regularly for the stability fault , Gradually improve stability ;

• Regularly publish the stability report of each business department , Set up a stability benchmark , Copy and promote excellent cases ;

• Improve the monitoring mechanism , Improve the failure rate of monitoring and detection .
  

Next , Let's look at the stability measurement data observation of this case .

Observation data 1 ： Through the measurement of fault related data, its asset loss is obtained , Including direct asset loss and potential asset loss . After we set our goals , Asset losses fell the following year 50%, The asset loss time is increased to 30 Within minutes . Of course , Stability improvement , It's a systematic project , We still have a lot to do ： In advance , We define the fault level （ for example , Level I requirements 5 Respond in minutes , Level IV requires response within half a day ）; In the matter , We also need to do emergency coordination for failures , How can we cooperate 、 How to solve the fault ; After the event , Scheme for improving the whole system , Improve the stability of future operation .

Observation data II ： Measure by analyzing the cause of the failure , Form improvement measures , The business stability improved in the following year 58%. We analyze the cause of the fault and the type of improvement measures , See what improvements we can make to avoid failures , At the same time, prevention and improvement .

Observation data III ： Through the analysis of fault discovery form , For the faults found on the user side , Reverse compensation monitoring logic , The next year, the fault occurs Increase of current rate 20%.

Observation data IV ： Through the regular stability report to the nine business divisions , Visualize quality . Give Way CTO、CEO See the overall quality , At the same time, let the experience of excellent departments be copied and popularized in the organization .

Scene two

Another scenario is the case of demand value measurement , Let's take a look at the pain points ：

• Pain points at the company level ： A lot of resources have been invested , But I don't know how to solve the demand blowout and whether to maximize the value ;

• Pain points on the technical side ： Technology output cannot be measured , It is not clear how many products have been made ;

• Pain points on the product side ： There are many sources of demand , Every business is said to be an urgent need , Unable to systematically design the product ;

• The pain points of technicians ： Write the code down , Do not understand the value of actual needs , As a result, the delivered version can not meet the expectation , Unable to get the sense of achievement brought by the job .

Based on these pain points , We designed the goal and steps of demand value measurement . The goal is Quantify the actual value of the demand , Prioritize requirements , Invest resources accurately , Maximize value ; Improve the recognition of the R & D side to the demand business objectives , And track the estimated and actual value deviation , Improve product design .

in addition , The realization steps of demand value degree are divided into three steps ：

First step , Sort out the demand value classification according to the current business objectives and past demands , Recommend from 「 The value of the business 」 as well as 「 Enterprise value 」 Two aspects of classification ;

The second step , In the project collaborative management tool configuration and traceability process , Here's the picture ：

The third step , The conformity of the value of regular re offer , Analyze the deviation problem , Observation effect .

Follow the steps above , First , Analyze the conformity of demand value . We investigated a business unit , A quarter has 46% The demand of did not meet the expectation ; secondly , We divided the whole business into four cross functional teams , Let the teams compare horizontally , Imitate each other 、 Study .

Last , Is the overall demand value compliance . such as , Strategic cooperation has reached 100%, Increasing revenue is also 100%. And improve the conversion rate 、 Risk avoidance and other indicators , Not meeting expected value of the demand .

According to customer feedback , The benefits of this method of analysis are It has aroused the enthusiasm of development . They are more willing to help the business achieve its goals through technical means , Because he 「 Know what it is , Know why 」, Can better understand this business .

「‍ Measure 」 Thinking and inspiration

in addition to , I would also like to share our thoughts on R & D measurement .

The first on the one hand, , Efficiency improvement landing , Standard tools first . First , Sort out the process specifications , Ensure that a series of consensus words are formed throughout the organization . for example , Awareness of goals 、 Uniform terminology 、 Serial connection of business and data flow 、 Unit of measure / Unification of measurement sources, etc ; secondly , Solidify process specifications through tools , Carry out measurement index embedding ; Last , By measuring the presentation layer , For each perspective 、 Experts in various fields present data they care about .

In the second , Measurement needs to be customized . For R & D performance measurement , The enterprise needs to combine the business objectives of the stage 、 Management honesty 、 Problems to be solved, etc , Tailored metrics .

The third aspect , Goodhart's law tells us ：「 Once an indicator changes , It is not a good measure .」 So metrics have to be weighed , The design of R & D metrics should also have a traction effect , Don't blindly pursue the improvement of numbers , It depends on whether the problem is really solved . also , Metrics should avoid being directly linked to performance . If performance measurement becomes a set goal , People tend to optimize it , Whatever the consequences , In this way, the measurement loses its meaning .

Scan the code to get the full version of the speech video and PPT