Most failing software builds don’t announce their failure. They drift quietly into ‘almost done’ territory and stay there for months. By the time the failure is visible to executives, the build is six weeks past being recoverable cheaply. Here are the four early signals we see across rescue engagements — the ones that show up before anyone says anything is wrong.

Signal 1: the demo always works, the staging never does

Healthy projects have a working staging environment that the team uses for everything: PR previews, customer demos, internal testing. Failing projects have a demo deploy that gets manually patched up before showings, and a staging environment that’s been broken for two weeks because ‘we’ll fix it after this sprint.’ If your team’s demo and staging are different things, that’s the signal.

What it actually means: the team has lost confidence in their CI/CD, started cherry-picking which work goes where, and is now delivering theatre rather than software. The fix isn’t ‘fix staging this week.’ It’s surfacing whatever made staging unmaintainable in the first place — usually a brittle integration, a flaky test suite, or an unowned piece of infra that everyone blames each other for.

Signal 2: you can’t say what ‘done’ means

Ask the project lead, in 30 seconds, what ‘shipped’ looks like. A healthy answer is concrete: ‘three named users complete onboarding, place an order, and the order shows in the back office.’ A failing-project answer is qualitative: ‘when the platform is stable enough’, ‘when the new flow is in’, ‘when the team is comfortable.’ Qualitative success criteria are an admission that nobody knows what done is.

What it actually means: scope has expanded incrementally over months, no one has had the conversation about cutting, and the team is now negotiating with itself. The fix is brutal: write the actual ship list, in 5 bullets, and accept that everything else is a separate project. We do this in week 1 of every rescue. It is uncomfortable for two days and then it is liberating.

Signal 3: bug reports stop arriving

This one is counterintuitive. You’d think no bugs is a good sign. It isn’t — it usually means nobody is using the system enough to find bugs, or that bug reports are no longer being filed because the team has stopped fixing them and people gave up. A healthy in-flight project has 5–15 bugs filed per week per customer. Zero is bad news.

What it actually means: the people who would file bugs (internal users, beta customers, QA) have lost the feedback loop. Either they’re not using it or they don’t believe filing helps. Restoring the loop usually involves both fixing existing bugs visibly (so reporters feel heard) and making the filing path frictionless. Until the loop returns, you don’t have visibility into what’s actually broken.

Signal 4: the team changes the metric

Watch for the phrase ‘we’re now tracking X instead.’ If a project was originally going to be measured by signups, then by activation, then by daily active users, then by retention, then by ‘qualitative engagement’, the metric isn’t moving — the team is moving the goalposts to find a metric that’s currently improving. This is a self-protective behaviour; nobody does it on purpose. But every rescue we’ve run has had at least one example of it.

What it actually means: the originally-promised outcome isn’t happening, the team knows, and they’re searching for a story that doesn’t say ‘we’re behind.’ The fix is to gently retire the new metrics and revisit the original one — not to punish, but to make the actual situation visible. Usually scope-cutting follows naturally once the gap is named.

What to do this week if you spotted yourself

If one of these signals is present, you have weeks not months. If two or more are present, you have days. The single highest-leverage move is naming the situation explicitly with the team — not as blame, as fact-finding. Most rescue triage weeks open with that conversation. The team is often relieved to have the unspoken thing said out loud.

If you’ve spotted yourself in any of the above, a triage week is the cheapest insurance you can buy. £8–12k, 5 working days, ends with a written ship plan you can take to your team or to us. Book the triage.