Owning software · 4 min
Finish, fix or rebuild: deciding what to do with software that stalled
Short answer: finish it if the foundations hold and what is missing is features. Fix it if it works for users but changing it is slow or risky. Rebuild only when the data model is wrong for what the business now does, or when the platform underneath is being retired. Most stalled software is a finish or a fix. A new team will often recommend a rebuild anyway, because people prefer code they wrote.
Why this decision goes wrong
The decision is usually made by whoever you ask first. A developer who wants the work suggests a rebuild. The original team, if they are still around, suggests finishing. Neither has an incentive to look at the question from your side, which is simply: what gets working software to users, for the least money and risk?
A rebuild also looks attractive because its cost is invisible at the start. You see a clean plan. You do not see the months of rediscovering every edge case the old system already handled.
The questions that settle it
1. Do real users depend on it today?
If people use it and pay for it, it holds knowledge nobody wrote down: the odd rule for a specific customer, the workaround for a payment provider's quirk. A rebuild has to rediscover all of it. That pushes strongly toward fixing.
If nobody uses it yet, the cost of starting again is lower, but so is the cost of finishing.
2. Is the data model right?
This is the one that matters most. Code can be improved one piece at a time. A data model that does not fit the business cannot. Signs it is wrong: every new feature needs a workaround, the same fact is stored in three places, or the product now serves several customers but was built for one.
A wrong data model is the strongest reason to rebuild, and even then the usual answer is to rebuild the core and keep the edges.
3. How far along is it, measured honestly?
"80% done" in a stalled project usually means the visible 80%. The remaining part is where the hard work sits: payments, permissions, error handling, the deploy, the admin tools someone has to use every day. Count what is left in working features, not in screens.
4. Can anyone other than the original author ship a change?
If a small change takes a week because everyone is scared of breaking something, that is a fix problem, not a rebuild problem. Tests around the fragile parts and a reliable deploy usually change how the whole codebase feels.
5. Is the platform underneath going away?
A framework version with no security updates, a hosting product being shut down, an app built on a no-code tool that cannot do what you now need. These are real reasons to move, and they are often mistaken for reasons to rewrite everything at once.
What each path usually looks like
- Finish: a fixed scope of the missing features plus the unglamorous parts, such as payments, admin and deploy. Weeks, not months, if the foundations are sound.
- Fix: tests around the fragile parts, a reliable deploy, the worst areas rewritten one at a time while the product keeps running. It can be a short project followed by ongoing ownership.
- Rebuild: the core rebuilt alongside the old system, data migrated, users moved over in stages. The most expensive path, and the one most likely to run over.
How to decide without guessing
Get someone who has no stake in the answer to read the system before anyone quotes for the work. That is the point of a paid, fixed-price check: the verdict is the product, so there is no incentive to steer you toward the biggest job.
Ours is the Reality Check, a fixed-price codebase audit: three to five working days, from €1,800, with a written keep, fix or rebuild verdict and what to do first. If you then start a project or a partnership with us within 60 days, the fee is credited. If you would rather take the document to another team, it is written so they can use it.
For what the check looks at in detail, see what a Reality Check covers.