How to manage Technical Debt
Today more people think about technical debt than 5 years ago. Because every project has a debt, but not everyone drives it. And even if they drive - they will just keep a snowball in most cases.
I want to share how I prefer to manage technical debt and what it gives to me in a result.
Definition and the problem
What Wikipedia tells us:
Technical debt (also known as design debt or code debt) is a concept in software development that reflects the implied cost of additional rework caused by choosing an easy solution now instead of using a better approach that would take longer.
If technical debt is not repaid, it can accumulate ‘interest’, making it harder to implement changes later on. Unaddressed technical debt increases software entropy.
In easy words, technical debt means a list of invalid solutions, which you or your team made by mistake. Do not forget - everyone can be wrong. So if you do not manage your debt - you will get the next list of problems.
- Demotivated team
- Have no ability to control project risks
- Increased software entropy
- Everyone would like to rewrite the whole project from scratch
Causes
List of causes can be very big, but I defined the most popular from my point of view.
- Requirements are still being defined during development
- Development starts before any design takes place
- Business pressure
- Lack of process or knowledge
- Last minute specification changes
- Poor technological leadership
- Lack of collaboration
- Outdated technologies
- Bad development practices
You noted that most of those causes related to a business or a client. And this is true - in most cases, technical debt is filled-up in collaboration with your manager. You should hear the next phrase.
Do it now because we really need.
And then you come to some compromise via increasing of your technical debt.
Disclaimer: At the end of the article you will find a link to the full list of root causes.
Technical debt and PoC
If you work on a Proof of Concept, do not worry about technical debt. Because someone will replace PoC with the more stable solution. The issue that everyone on a project should be on a same page.
Popular solution
In most cases, developers think it is enough to write TODO item in their code. But imagine if you have 40 repositories for micro-services on a big enterprise. How will you collect all TODO-s? Scan the sources? Okay, but how will you track progress? If you do not see how much you have done - you do not feel happy.
Here I would like to lead to my way of technical debt management.
More precise solution
I propose to use the next structure as a table. You can extend the structure with additional columns, but please - do not shrink.
- Number
- Issue
- Cause
- Solution
- How to prevent?
- Initiated
- Solved
- Status
- Assignee
Seems big, but let me dive in and describe every point in details. Except for Number - an auto-incremental stuff.
Issue - Cause - Solution - Prevention
The issue used for code-name, which will help you find and identify an item. But any issue has a root cause, and it is critical to define the cause.
When you fill the Cause column answer next questions: Why/How/When an issue occurred? Questions will help you define a root cause and understand who made a mistake and why.
When you know the root cause - you can fix an issue. And it is the best time to describe your solution. You can brainstorm with a team and write the solution, so everyone will be on the same page.
And do not forget - in future you need to prevent that issue. If you have the same issue over 3 times - you have a problem in your development process and something is going wrong.
Initiated vs Solved
Initiated means that you found or created an issue in some period. You can write the date, but I prefer to write the number of a Sprint or Iteration. In a such way, you can analyze a table and see a correlation between rush-sprint and increased technical debt.
And also do not forget to write when you solved an issue. If you do things right - you can build such chart.
You can use resolution chart to define when you made a lot of invalid descisions. Also, you can find out that you forgot to fix debt for a long time. Non-stacked area chart is best for that case. For vertical axis - use amount of total/resolved issues per sprint or date.
Status and Assignee
You can use Jira or any other task tracking system or do not use it at all. But please - do not remove a completed item from the table. Otherwise - you will not be able to track a progress and amount of completed stuff. Also do not forget that you wrote how to prevent that issue in future.
Example of an item
Here I provide an example of an item in technical debt.
| Column | Value |
|---|---|
| Number | 13 |
| Issue | Outdated README file |
| Cause | After creation of landing page, there were changes in a folder structure |
| Solution | Update the file |
| How to prevent? | Update the README file right with changes in structure in the same Pull Request |
| Initiated | 21 |
| Solved | 25 |
| Status | DONE |
| Assignee | Aleksey Pastuhov |
Here I want to note that you can fix an issue in time, it is simple prioritization.
Summary
Remember that you should monitor your debt iteratively, and take tasks into your sprints to get things done.
This is how I prefer to keep and manage technical debt. If you do right, you will have positive results:
- A motivated team
- You will be able to control technical risks
- If there are no requirements or design - you have a lot of stuff to do
- And you will be able to debate to your manager or a client that you need a sprint for code stabilization
Share your feedback and let me know how you used to manage technical debt. Send me a message or tweet!
P.S.: Do not forget about prioritization! Just add “Priority” column, if it's required in your case.