Incident Count per 1k Lines of Code

What is this metric?

Amount of incidents per 1,000 lines of code.

Why is it important?

1. Defect drill-down analysis to inform the development of design and code review strategies and to improve the internal QA process
2. Assist teams to locate projects/modules with higher defect severity and density, and clean up technical debts
3. Identify good/to-be-improved practices that affect defect count or defect rate, to reduce the number of future defects

Which dashboard(s) does it exist in

N/A

How is it calculated?

The number of incidents divided by total accumulated lines of code (additions + deletions) in the given data range.

Data Sources Required

• issues collected from Jira, GitHub, TAPD or PagerDuty.
• commits collected from GitHub, GitLab or BitBucket.

Data Transformation Required

This metric relies on the 'type-incident' configuration in Jira, GitHub or TAPD's transformation rules while adding/editing a blueprint. This configuration tells DevLake what issues are incidents.

SQL Queries

If you want to measure the monthly trend of Incidents per 1k lines of code in the screenshot below, please run the following SQL in Grafana.

with _line_of_code as (
    select 
      DATE_ADD(date(authored_date), INTERVAL -DAY(date(authored_date))+1 DAY) as time,
      sum(additions + deletions) as line_count
    from 
      commits
    where 
      message not like 'Merge%'
      and $__timeFilter(authored_date)
    group by 1
),


_incident_count as(
  select 
    DATE_ADD(date(created_date), INTERVAL -DAY(date(created_date))+1 DAY) as time,
    count(*) as incident_count
  from issues i
  where 
    type = 'INCIDENT'
    and $__timeFilter(created_date)
  group by 1
),


_incident_count_per_1k_loc as(
  select 
    loc.time,
    1.0 * ic.incident_count / loc.line_count * 1000 as incident_count_per_1k_loc
  from 
    _line_of_code loc
    left join _incident_count ic on ic.time = loc.time
  where
    ic.incident_count is not null 
    and loc.line_count is not null 
    and loc.line_count != 0
)

select 
  date_format(time,'%M %Y') as month,
  incident_count_per_1k_loc as 'Incident Count per 1000 Lines of Code'
from _incident_count_per_1k_loc 
order by time;

How to improve?

1. From the project or team dimension, observe the statistics on the total number of defects, the distribution of the number of defects in each severity level/type/owner, the cumulative trend of defects, and the change trend of the defect rate in thousands of lines, etc.
2. From version cycle dimension, observe the statistics on the cumulative trend of the number of defects/defect rate, which can be used to determine whether the growth rate of defects is slowing down, showing a flat convergence trend, and is an important reference for judging the stability of software version quality
3. From the time dimension, analyze the trend of the number of test defects, defect rate to locate the key items/key points
4. Evaluate whether the software quality and test plan are reasonable by referring to CMMI standard values