Chip Cullen

How to Deal With Large Pieces of Technical Debt

At my work we've successfully navigated a few very large chunks of technical debt. One instance was removing an old dependency, but there has also been fixing issues picked up by linters, and converting to Typescript large pieces of applications. Having seen this happen, but also having seen instances where technical debt hasn't been addressed, I've come to learn this:

The smaller the individual pieces of work, the greater the chance that a large piece of technical debt will be addressed.

And I mean small - like, atomic level. One ticket = one file. Really small.

Don't get me wrong - this will mean some solving some problems up front. But I cannot stress enough how critical this is to success.

Problem one: how can this improvement be incremental?

The biggest obstacle to technical debt getting addressed is if the change is huge and is done all at once. If at all humanly possible, try to figure out how to incorporate the kinds of changes necessary on one file at a time without affecting other pieces of the application.

For example, converting one file to TypeScript means other, non-TS-javascript files can still import from it. Take it one file at a time.

There are times when a massive change is unavoidable. Like switching from one major version of a language to the next (e.g. Python 2 to Python 3). The best advice I can give in a situation like this is to make sure you have good tests covering your critical functionality in place. And, ideally, those tests are observing your application from another point in the stack (e.g. frontend e2e tests that verify a backend upgrade doesn't break anything).

Problem two: identify each tiny piece of work, and turn it into an actionable task

If you're a manager, or anyone else responsible for creating tickets, I'm not gonna lie - this part is a slog. Figuring out how to break work down is a skill unto itself, and then there is the task of creating all the tickets. This kind of sweat up front will pay off huge dividends in terms of the work actually getting done.

If at all possible, investigate ways to bulk-create tickets from a CSV. That is a game changer for dealing with these kinds of issues.

Problem three: get team buy in on how it's getting addressed, and the overhead involved

For the team involved, talk through what these atomic level pieces of work will mean. Things like: there will be a lot of PR's, so get ready for a lot of notifications. How do you want to be notified? What is the understood testing procedure, etc?

Benefits to atomic level pieces of technical debt