In Jira, one can attach the label of “priority” to an issue to define how quickly the issue should be completed. How companies set this priority level varies, but in general it falls into one of three categories: Low, Medium, and High. Medium priority is often the default label and implies that the work can be completed in a standard development (dev) cycle. Low priority implies that the work is less important than that in the current dev cycle. And high priority implies that the work should take precedence over work in the standard dev cycle.
These priority labels suggest a question seemingly too simple to ask: Does the higher priority of an issue actually correlate with a faster completion time?
We explored this question across 270,000 issues for 50 companies. For all companies, we only considered resolved issues (i.e., issues whose final state was “Done”) and we defined “issue cycle time” as the number of days it takes an issue to go from “Created” to “Done.” In the plots below we show the distribution of issue cycle times across this aggregated collection of companies for each issue priority.
Together with these plots we can compile basic descriptive statistics of the respective distributions. We have:
Some comments about the meaning of these results.
- Long Tailed Distributions: First, issue cycle time is a prototypical long-tailed distribution which means that unlike normal distributions that define things like human height or shoe size, there is a non-negligible likelihood that an issue cycle time will be found far from the mean of the distribution. Less formally, this implies that a small but non-zero proportion of issues will just take a really long time and such issues should not be considered outliers but are in fact real parts of the distribution.
- Low Priority for Low Priority: We note that most issues are medium priority, followed by high priority, and an almost negligible number of issues are low priority. There is a good reason for this. Low priority issues are so low on the totem pole of importance that it seems to be a low priority to assign them as such.
- Priority makes little difference: Finally, and most relevant to our initial question, we see that the distribution for high priority issues does not have a smaller mean completion time than the distribution for medium priority issues. This leads us to the unexpected conclusion that medium priority issues are completed, on average, faster than high priority ones. How can we better understand this result?
Due to a non-negligible portion of long-cycle-time issues, high priority issues, on average, take almost a day longer than medium priority issues to be resolved and two days longer than low priority issues.
The Impact of Varying Cycle Times & Jira Maintenance/Management
The above data is from many different companies with disparate processes and cycle times. For one company, taking a week to complete an issue could be the norm while at another company that could be the sign of poor Jira hygiene or project management. Also, one company could be quick to label an important issue as “high priority” while another company might find ways outside of Jira to convey issue importance.
In essence, different companies have different standard cycle times and different tendencies to attach high priority labels to issues. We can affirm this by looking at distributions for the average cycle time and the ratio between total the number of high priority issues and the total number of issues:
For some companies, a few days is a typical cycle time for resolving issues, while for other companies two weeks is more standard. Few companies have average cycle times more than three weeks. Similarly, some companies rarely categorize an issue as high priority, while for other companies all issues are high priority.
Across organizations, average issue cycle times vary from a few days to a month, and while high priority issues are rare for some organizations, for others almost all issues are high priority.
Therefore, our above result that high priority issues take, on average, more time than low priority issues might be an artifact of having companies with many issues and a higher than typical tendency to label an issue as high priority having a higher than typical cycle time. Maybe within those companies high priority issues are still completed on the time scale or even faster than low priority ones?
So to see if high priority issues do indeed take longer to resolve (at least according to Jira if not the actual project flow) than medium priority issues, one would need to disaggregate this data across companies.
To do so, we created the plot below which is a histogram of the difference between the mean cycle times for high priority and low priority issues for all companies. For example, the bar to the right of zero signifies that 37% of the companies have an average cycle time for high priority issues that is a few days greater than the average cycle time for medium priority issues.
Thus the result found in the aggregate plot still stands but with a bit more precision: For some companies, medium priority issues have shorter cycle times (i.e., are completed faster) on average than high priority issues.
There are more companies where high priority issues are completed at the same rate or even slower than medium priority issues than there are companies where high priority issues are completed faster than medium priority issues.
Even this more precise result is a bit unbelievable, but there is an explanation for it.
Jira is the only well-defined signal we have to measure the time from project start to project completion. But it is an imperfect signal because it requires project managers to move tickets in a process that runs parallel to the process by which the project is completed. Sometimes these parallel tracks can get out of sync.
So what the above result suggests is a difference in attention to Jira hygiene for different priorities of issues. For high-priority issues the main focus is on completing the work that the issue represents. This is because such issues may be associated with application crashes and are resolved quickly after which the issue management is seen as less important. Consequently a high priority issue can linger in an intermediate state long after the work associated with it has been completed. The Jira cycles of medium priority issues, which are perhaps completed under less pressure, can be better managed with a focus on properly tracking when the issue was completed. Low priority issues (for the few that exist) are perhaps quick fixes that are both easier to resolve than medium priority issues and have the least amount of time pressure for completion.
There are many possible implications of these findings, and hopefully they prompt further discussions with your team as to what your issue classification in Jira might be telling you. In order to help facilitate these conversations, we’ve summarized a few key questions based on our findings that you may wish to raise with your team:
Questions to ask your team:
- Are we usefully labeling and treating “Low priority” issues as such?
- How long do we expect our issues to take on average?
- Are we categorizing too many issues as high priority?
- Where does our organization or team sit in the distribution of how average cycle time for high priority issues relates to average cycle time for medium priority issues?
- Are we forgetting to move high priority issues to done after they’re completed?