Fortnox的数据驱动软件工程实践.pdf

1、Henrik Edegrd/Javier Gonzalez Huerta-2024-11-20Data-driven software engineering at Fortnox 01Henrik EdegrdDevelopment ManagerDeveloper ExperienceJuni 202301Fun Fact:#:Update sales_table set firstname=henrik,lastname=edegard,street=myroad;#:Affected rows:1 500 000#:PanicJuni 202302630 00002Developer

2、1:-Im tired of npm update02Developer 2:-I know a tool02Developer 1:-Wow,this solved my problem02Developer 1 to Manager 1-Check this out,we all need this02Manager 1 to“Man-in-charge”-We should all do this02“Man-in-charge”to rest of Managers:-Please do this02“Man-in-charge”to rest of Managers:-Please

3、do this02Random Developer:-I get swamped with PR:s02Random Developer 2:-Hey,I made this script for chrome to hide your Renovate PR:s03So.?Did anyone ask for data or questioned whyA developer initiated projects often misses looking at data or defining metrics for success.03DataStarting digging into d

4、ata.What goal should we have with a project like this?Reduced number of CVE:s in deps?More time spent on business logic?Lets do a educated guess03The data50 teams500 developers500 Repositories03The dataRenovate Pull Requests in 2 years25 000 opened21 000 merged4 000 DeclinedAvg time to merge:12 days

5、(at 8h per day)03The dataWith 21 000 mergedat 3 approves per PREquals to 63 000 Approvesat 10 minutes of work per approveTADA:630 000 minutes of developer time.(10 500h/1 312 working days/5,6 years)03The data5.6 years of work over 24 monthsAre 2.8 full time employeesAnd what was the outcome?03The da

6、taProbably more up to date dependenciesMore secure?Probably,but did you measure it?03The dataDid we get any better?-How can you decide if you did not have any metrics?Tweaks that can be made-group patch/minor to one PR-reduce when to run(once a week)03Not the