How to Rescue a Failing Software Project
Every rescue engagement starts the same way: a system in production, a team (or a departed vendor) that can no longer explain it, and a business that depends on it anyway. The instinct is to start fixing immediately. The instinct is wrong.
Step one: diagnose before touching anything
The first phase is read-only: profiling, query analysis, log review, infrastructure inventory, and a risk register that ranks what can actually hurt you: data loss beats slow pages, silent corruption beats loud crashes. A fixed-price diagnostic ($5,000–$12,000, one to two weeks) produces a written findings document and a prioritized plan. You pay a small fee to know exactly how bad it is before committing to anything, and the document is yours even if you take it to another firm.
Step two: stabilize in the right order
- Backups first. Before any change, verify a restore actually works. Rescues have been turned into disasters by fixing a system whose only backup was fictional.
- Observability second. Install monitoring and error tracking (Datadog, Sentry) so every subsequent change has a visible effect. You cannot fix what you cannot measure, and you cannot prove a fix without a baseline.
- Bleeding third. Stop active data corruption and security exposure before performance, and performance before polish.
Step three: fix in tranches, not a rewrite
The most expensive sentence in software is "let's just rebuild it." Rewrites discard years of embedded edge-case knowledge and routinely fail. Rescues work as bounded tranches: each with a measurable before/after (p95 latency, error rate, infrastructure cost), each independently reversible, each billed with a cap. Value is proven, not promised, and you can stop at any tranche with the system better than you found it.
What a rescue costs
| Phase | Range | Duration | Output |
|---|---|---|---|
| Diagnostic | $5,000–$12,000 | 1–2 weeks | Findings document, risk register, prioritized plan |
| Typical remediation | $20,000–$60,000 | 2–10 weeks | Tranches with measured before/after |
| Severe / legacy systems | $75,000+ | Months, staged | Undocumented or abandoned systems carry a discovery premium |
Choosing a rescue partner
Ask two questions. "What will you do in the first week?" (right answer: measure, not code). And "how will we know it worked?" (right answer: specific numbers, named up front). A firm that answers with technology names instead of measurements is auditioning to be your next rescue.
Frequently asked questions
Should we rewrite or fix the existing system?
Fix, in almost every case. A rewrite discards the edge-case knowledge encoded in the current system and takes twice as long as promised. Rewrites earn their place only when the platform itself is terminal, and that conclusion should come out of a diagnostic, not a sales call.
Our developer disappeared with the credentials. Where do we start?
Ownership recovery is step zero: domain registrar, hosting, DNS, repositories, third-party accounts. Then the standard diagnostic. This is common, unglamorous, and usually takes days, not weeks.
How fast can performance problems be fixed?
Often surprisingly fast once measured: the majority of severe slowdowns trace to a handful of database queries or missing indexes. The diagnostic tells you whether yours is that case or a deeper architectural issue.
Talk to the people who do this work
Ashton Group provides performance tuning & monitoring for organizations across Maryland, DC, Virginia, and beyond. Tell us about your project and we'll reply with a candid assessment within one business day.
Email hello@ashtongroup.org About this service