# Trap 8 - The watermelon

> Green outside, red inside. Why good people report green on troubled programs: ten habits that grow a watermelon, and how to make the truth cheap to say.

Source: https://www.bluewise.fr/why-programs-fail-the-watermelon-trap  
Author: Marc Morel  
Published: 2026-10-15  
Category: Programs  
Series: Why programs fail, part 9

## The emperor's new clothes

Two swindlers arrive in a city and promise the emperor the finest cloth in the world, with one magical property: it is invisible to anyone unfit for their job, or hopelessly stupid.

The emperor sends his most trusted minister to check on the weavers. The minister sees empty looms. He panics quietly, then praises the colors and the patterns. The next official does the same, and so does the one after him. Each sees nothing. Each reports a masterpiece.

On the day of the great procession, the emperor walks through the streets in his new clothes. The crowd cheers, because nobody wants to be the one who can't see them. Then a child says what everyone can see: he isn't wearing anything at all.

Hans Christian Andersen wrote this story in 1837. Nearly two centuries later, it still plays every month in steering committees. Only now, the clothes are green.

## Green outside, red inside

You can't tell a ripe watermelon by looking at it. The shell is the same smooth green whether the inside is perfect or rotten. You only know when you cut it open. Many programs work exactly the same way.

For eighteen months, the steering committee sees green. The dashboard is green, the milestones are green, the budget line is a reassuring shade of amber that everyone agrees is "under control." Then, six weeks before go-live, someone finally runs the end-to-end test. Overnight, the program goes red. Not amber. Deep, irreversible red.

The sponsor asks the question every sponsor asks in that moment: *how did we not see this coming?*

The honest answer is uncomfortable. Someone did see it coming. Several people, probably. The truth just never made it up to the room where decisions are made.

Let me be clear about one thing: watermelon programs are rarely built by liars. In 25 years, I have met very few people who knowingly reported green on a red program. What I have met, many times, are good people caught in a system that turns honest doubts into green dots.

It happens one reasonable step at a time.

The team lead thinks: *we're behind, but we'll catch up next sprint.* That is optimism, and every builder needs some. The project manager thinks: *three teams are amber, two are green, so overall we're green-ish.* That is aggregation. The program director thinks: *if I show red now, I'll spend a month in audits instead of fixing the problem.* That is self-protection, and it is often rational.

Researchers have a name for this: the *mum effect*, our deep reluctance to be the bearer of bad news. [Research on runaway IT projects](https://jmis-web.org/articles/885) has studied it for over twenty years. People rarely hide problems out of malice. They stay quiet because speaking up feels risky, and the benefit feels uncertain.

**That is the watermelon trap: a reporting system where green is the safest color to say, whatever the color of the truth.**

## The ten faces of the trap

The green shell is never painted in one stroke. In my experience it grows from ten habits, in three families. Most watermelon programs have four or five of them.

### How the status is produced

- **The status is an opinion.** The color is chosen by the person accountable for it, based on how they feel about next month. Nobody asks what evidence sits behind it.
- **One shade lighter per level.** The developer says red, the team lead reports amber, the program reports green-with-watchpoints. Each step softens the truth a little, with good intentions, until the top of the chain sees a different program from the bottom.
- **The catch-up plan that can't catch up.** Every slip comes with a recovery plan, and every recovery plan assumes the team will suddenly go faster: fewer bugs, quicker decisions, no sick days. The status stays green on the strength of a velocity nobody has ever seen.
- **The silent rebaseline.** When a milestone slips, the plan is quietly redrawn. The program is always "on plan." It just isn't the same plan as last quarter.

### What gets measured

- **Activity instead of outcomes.** Percentage of tasks complete, story points burned, documents delivered, hours spent. All of it can be green while nothing works end to end.
- **The 90% syndrome.** Every workstream is 90% done, and has been for three months. The last 10%, integration, data, performance, real users, is where the real work hides.
- **Goodhart's law.** When a measure becomes a target, it stops being a good measure. Once the milestone dates are what gets judged, milestones start being declared "achieved with reservations."

### How the room reacts

- **Red is punished.** A red status triggers audits, extra reporting and senior visitors. The team learns that red costs more than the problem itself, so red is avoided until it can no longer be.
- **The steering committee as theater.** Sixty slides, a packed agenda, no decisions. The meeting is designed to reassure, not to decide.
- **The vendor's green.** The integrator reports green on its contractual scope. Every supplier is green on its own part, and nobody owns the color of the whole.

## Late red leaves no options

Here is the real cost of a watermelon, and it isn't the bad news itself. Programs always meet bad news. The cost is *when* it arrives.

A problem spotted in month three comes with a full menu of options. You can cut scope, change the approach, reassign people, renegotiate with a vendor, or even stop the program cheaply. The same problem discovered in month sixteen comes with a much shorter menu: more money, more time, or both. Most of the other doors closed quietly while the dashboard was green.

And there is a second, slower cost: trust. Once a sponsor has been surprised by a deep red, every green after that is suspected. Steering committees become interrogations. Reporting doubles. The program slows down even further, because nobody believes the colors anymore.

The cruel irony: the watermelon was meant to protect the program. In the end it removes the one thing that could have saved it, time to react.

## Early signals

You can often feel a watermelon before you can prove it. Watch for these:

- The status hasn't changed color in six months, in either direction.
- Milestones are "achieved" but nobody can show them working.
- Percentages of completion climb slowly near 90% and stay there.
- The plan has a new baseline every quarter, and each recovery plan assumes the team will suddenly go faster.
- Team members are more worried in the corridor than in the meeting.
- The steering committee ends without a single decision recorded.
- Questions get answered with "we're working on it" rather than a date and a name.
- Every supplier is green, and the integration is not.
- Someone who raised a concern last quarter is no longer invited.

## Make the truth cheap to say

The answer isn't more reporting. Watermelon programs usually have plenty. The answer is to **make the truth cheap to say and hard to hide.**

1. **Thank the first red.** The sponsor's reaction to the first bad news sets the culture for the whole program. Say thank you, out loud, in front of everyone. Then ask what help is needed, not who is to blame.
2. **Evidence, not opinions.** Every status comes with proof: a demo of working software, a test result, a production metric. If nothing can be shown, the status is amber at best, whatever anyone feels.
3. **Measure outcomes, not activity.** Track what works end to end: flows running in a real environment, data migrated and reconciled, users onboarded. Leave percentages of tasks complete to the team's own boards.
4. **Show the trend and the forecast.** A burn-up chart against the original baseline tells more than a color. Where are we, where were we supposed to be, and when will we really land?
5. **Generate status from the work itself.** Pipelines, tickets, test runs and incidents already know the truth. A dashboard built from them is harder to soften than a slide built for the meeting.
6. **Amber needs a date and a name.** "Amber" without a recovery date and an owner is just a polite red. Make that rule explicit, and check that the recovery date rests on the team's measured pace, not a hoped-for one.
7. **Ask the watermelon question.** At every steering committee, ask each lead: *what would turn this red, and how close are we?* It invites people to speak about risks without having to declare failure.
8. **Bring fresh eyes.** A pre-mortem at each phase gate, or a short independent review by someone outside the reporting chain. Outsiders have no shade to protect.
9. **Steer, don't watch.** Every steering committee ends with recorded decisions: scope cut, people moved, risk accepted, or nothing to decide. Ten slides that lead to a decision beat sixty that lead to reassurance.

Done well, this doesn't feel like surveillance. It feels like relief. Teams stop carrying bad news alone, and sponsors stop being surprised.

## What to avoid

- Averaging statuses: one red workstream makes the program at least amber.
- Rebaselining without showing the original plan next to the new one.
- Turning red into a sanction: audits, extra reporting, public questioning.
- Reporting percentages of completion to the steering committee.
- Letting each supplier report only on its own scope.
- Steering committees that inform but never decide.
- Shooting, or quietly sidelining, the messenger.

## The AI twist

AI cuts both ways here. On the bad side, a polished status deck now takes minutes. Confident summaries, clean charts, reassuring wording: the shell of the watermelon has never been so smooth or so cheap to produce. A beautiful report no longer proves that anyone looked closely at the work.

On the good side, AI can read what slides never show. Tickets that keep reopening, tests that keep failing, pull requests waiting for days, incidents in the test environments. Pointed at the raw signals of the work, an assistant can flag the drift long before anyone dares to change a color. Used that way, AI doesn't write the status. It cuts the watermelon open.

## Red early is a gift

The best programs I have worked on were not the ones that stayed green. They were the ones where someone could walk into the steering committee, say "we have a problem," and walk out with help instead of blame. Their reds came early, so they stayed small.

So next time you chair or attend a steering committee, try one gentle question: *if this program were secretly red, how would we know?* If nobody has a good answer, you may be looking at a watermelon.

Have you lived through a program that was green until the very last moment, or one where an early red saved the day? I would genuinely love to hear your story.

*All situations described here are composites, drawn from many programs over 25 years.*