Missed dates, vanishing demos, fixes that break other things, a developer who stops replying. Here is how to tell a rough patch from a failing project, and what to do before more money goes in.
Short answer: a software project is failing when dates slip without a new plan, demos stop, every fix breaks something else, the bug list keeps growing, or you cannot get the code and hosting in your own name. Secure access and backups first, freeze new features, then get a written audit before deciding to rescue or rebuild.
Most failing projects do not fail on one day. They drift. A deadline moves by two weeks, then by a month. The weekly demo becomes a weekly message. Then one day the developer stops replying, and you realise nobody else knows how the system works. This guide is for the owner or manager in the middle of that drift: how to read the signs, what to do this week, and how to decide whether the software is worth saving.
What are the warning signs a software project is failing?
One sign on its own can be a bad week. Two or three together, lasting more than a month, are a pattern. Use this table as a checklist, and ask the question in the last column this week.
| Warning sign | What it usually means | What to ask this week |
|---|---|---|
| Dates keep slipping, with no new plan | Nobody has re-estimated the remaining work | "What is left, and what is the new date for each piece?" |
| Demos have turned into status messages | There is little working software to show | "Can you show it running on a staging link on Friday?" |
| Every fix breaks something else | No automated tests, tangled code, or both | "How do you check a fix did not break anything else?" |
| The bug list grows faster than it shrinks | The team is firefighting, not finishing | "How many bugs were opened and closed this month?" |
| You cannot access the code, server or domain | The project is effectively owned by the developer | "Please add our company account as owner today." |
| Changes go straight to the live system | No staging environment and no rollback plan | "Where do you test changes before customers see them?" |
| Scope keeps growing without a written record | Requests are agreed in chats and never priced | "Can we keep a written change list with dates and costs?" |
| Most of the budget is spent, core features are not | Effort went into the wrong parts first | "Which core user journeys work end to end today?" |
| Nobody can explain how to deploy or run it | Knowledge sits in one person's head | "Is there a README that lets another developer run it?" |
| Passwords and keys are shared in chats | A security risk that grows every week | "Where are our credentials stored, and who has them?" |
Is it a rough patch or a failing project?
Every real project has rough patches. The difference is whether the team can still explain where it stands. In a rough patch, the developer can tell you what went wrong, what is left and when it will be done, and the next demo shows progress. In a failing project, answers get vaguer each week, estimates are repeated rather than revised, and the same complaints come back after they were "fixed".
A useful test: ask for a one-page written status listing what is finished, what is half-built and what has not started. A healthy team can write it in a day. If nobody can, the project has lost its map, and that is the real problem to solve first.
What to do in the first week
Before anyone writes another line of code, protect what you already have. These steps are cheap, and skipping them is how a delayed project becomes a lost one.
- Get access in your company's name. Code repository, server or hosting panel, domain registrar, database, app store accounts, payment gateway and email service. Ask to be added as owner, not as a guest.
- Take a full backup. Code and data, stored somewhere you control, dated. Do this before any new developer touches anything.
- Freeze new features. Adding features to an unstable system makes every later fix more expensive.
- Write down your top five complaints. For each: what happens, the steps to reproduce it, and how often. This becomes the test list for any fix.
- Ask the current team for a written status and the latest code. Keep it polite and factual; you may still need their help with handover.
- Get an independent audit. A second team reading the code, the data and the hosting, and writing down what they find, before you decide what to pay for next.
Do not pay for "a quick fix" first. A fix to a system nobody has audited is a guess. If it fails, you have spent money and still do not know what you own.
Rescue or rebuild: how to decide
Rescue means stabilising the existing code, fixing it and finishing it. Rebuild means starting again, usually reusing the data and what you learned. Neither is automatically right; the evidence in the code decides.
| Rescue is usually right when | Rebuild is usually right when |
|---|---|
| The core user journeys work, even if roughly | The code cannot be run or deployed reliably |
| The technology is current and supported | The framework or language version is no longer supported |
| Problems are bugs, gaps and missing tests | The data model does not match how the business works |
| The data structure is sound | Each repair costs more than writing that part again |
| Customers already rely on it daily | Very little of it is in real use yet |
A partial path also exists: keep the system running and replace it one piece at a time. We explain that approach in modernising a legacy system without a two-year rewrite. If the project is an off-the-shelf tool bent out of shape, custom ERP vs off-the-shelf covers when to stop customising.
What should a software rescue audit cover?
An audit is only useful if it is written, specific and backed by evidence. Ask for these items:
- Inventory: where the code, hosting, database, domain and third-party services live, and who owns each account.
- What works and what does not: every main page, screen and API tested, with steps to reproduce each failure.
- Security: keys or passwords leaked in the code or its history, open database rules, outdated packages with known issues, missing backups.
- Risks ranked: what must be fixed now, soon, or later.
- Rescue vs rebuild: a recommendation with the evidence for it, and both options priced if a rebuild is on the table.
- A phased plan: stabilise, fix, finish, hand over, with a price and an outcome for each phase.
Be cautious of an audit that blames the previous developer at length and says little about the code. Blame does not get software working again.
How to stop the next project failing
Most failures trace back to a few habits that are easy to put in the contract from day one:
- Every account (code, hosting, domain, app store) registered in your company's name from the start.
- A working demo on a staging link every week or two, not just a status message.
- Small releases, each with a way to roll back.
- A written change list: every new request with a date and a price before work starts.
- A README that lets any competent developer run the project without the original team.
Our guide to twelve questions to ask before hiring a digital agency turns these into questions you can ask before signing.
How Kredosys rescues a project
Our software rescue service starts read-only: we back up all code and data before our first change, then write an audit, usually within 48 hours of getting access. It lists what exists, what works, what is broken and what is unsafe, with a fixed-price plan in phases. No code changes until you accept it. Code, hosting, domain and app store access end up in your name, and Himanshu Soni stays accountable from the audit to the handover. For new builds rather than rescues, see our custom software work.
Frequently asked questions
What are the signs a software project is failing?
The clearest signs are dates that keep slipping without a new plan, working demos replaced by status updates, fixes that break other features, a bug list that grows faster than it shrinks, and code or hosting you cannot access in your own name. Two or more of these together usually mean the project needs an independent audit.
Can a failing software project be saved?
Often, yes. Many failing projects have a sound core with missing pieces, weak testing or poor release habits, and those can be stabilised and finished. Some have foundations so weak that each repair costs more than starting again. Only a written audit of the actual code, data and hosting can tell you which case you are in.
What should I do if my software developer stops responding?
First secure access: the code repository, server or hosting, domain, database, app store and payment accounts, moved into accounts your business owns. Take a full backup of code and data. Then send a written request for the latest code and a status note, and get an independent audit before paying anyone to write new features.
Should I rescue my software or rebuild it from scratch?
Rescue when the core works, the stack is current and the problems are bugs, gaps and missing tests. Rebuild when the code cannot be run or deployed reliably, the data model is wrong for the business, or the technology is no longer supported. A good audit prices both options side by side so you can compare.
How much does it cost to rescue a software project?
It depends on what the audit finds: how much works, how much is missing and how risky the code is. Be wary of anyone who quotes a rescue before looking at the code. Kredosys starts with a written audit and then quotes a fixed price for each phase, so you agree to one phase at a time.