# Trap 5 - The decision

> Programs rarely die of bad decisions. They die of decisions that never quite get made.

Source: https://www.bluewise.fr/why-programs-fail-the-decision-trap  
Author: Marc Morel  
Published: 2026-10-15  
Category: Programs  
Series: Why programs fail, part 6

## The road to Abilene

It is a Sunday afternoon in July, in Coleman, Texas, and the thermometer says 104 degrees. A family sits on the back porch with cold lemonade and a game of dominoes. A perfect afternoon.

Then the father-in-law says: "Let's drive to Abilene and have dinner at the cafeteria." Abilene is more than fifty miles away, and the car has no air conditioning. His son-in-law says it sounds great. His wife says she'd love to go. Her mother says she hasn't been in a long time.

Four hours later they are back, dusty and exhausted, after a mediocre dinner. Someone says it was a great trip, and the truth finally comes out: nobody wanted to go. Not even the father-in-law, who only suggested it because he thought the others were bored.

Jerry B. Harvey told this story in 1974 and called it the Abilene paradox: a group can make a decision that every one of its members privately disagrees with.

I have been to Abilene many times. Never in Texas. Always in a steering committee.

## The hidden critical path

Every program has a plan full of tasks: build this, test that, migrate the rest. What the plan rarely shows is the other critical path, the one made of decisions. Which vendor? Which scope for release one? Do we migrate the history or not? Who owns the customer data?

Each of these questions blocks real work. And unlike tasks, decisions have no estimate, no owner on the Gantt chart and no alarm when they are late.

The people involved are rarely indecisive by nature. Most sponsors and program directors I've worked with make dozens of good decisions a week. The trap is in the system around them: too many people consulted, nobody clearly accountable, every decision treated as equally serious, and every decision open to reopening.

**That is the decision trap: a program that moves at the speed of its slowest decision, without ever measuring that speed.**

## The ten faces of the trap

Ten shapes, in three families: who decides, how we decide, and whether the decision holds. Most stalled programs show three or four at once.

### Who decides

- **No owner.** Everyone is consulted, nobody decides. The question circulates between committees until the deadline decides for everyone.
- **The HiPPO.** The highest-paid person's opinion settles the matter, whatever the evidence. Teams learn to guess what the boss thinks instead of finding out what is true.
- **The trip to Abilene.** Everyone agrees out loud and disagrees in private. The decision is unanimous, and nobody believes in it.

### How we decide

- **Too many gates.** Architecture board, security review, budget committee, steering committee, each meeting monthly. Every gate adds weeks of waiting for a few minutes of decision.
- **Bikeshedding.** C. Northcote Parkinson described a committee that approves a nuclear plant in minutes and then debates the bicycle shed for an hour, because everyone understands a bike shed. Programs spend hours on logo colors and minutes on the data model.
- **Every decision is a one-way door.** Reversible choices, like a library, a screen layout or a test approach, get the same scrutiny as irreversible ones, like a core vendor, a data model or a contract. So the small ones are slow, and the big ones get rushed to catch up.
- **Waiting for certainty.** One more study, one more benchmark, one more opinion. Information keeps arriving, and certainty never does.

### Whether it holds

- **The boomerang.** A decision taken in March comes back in April because someone missed the meeting, or simply didn't like it. Nothing is ever really decided.
- **Meeting amnesia.** The decision was made, but nobody wrote it down. Six weeks later, three people remember three different versions.
- **The silent veto.** Everyone agrees in the room, and then nothing happens. The decision is respected on paper and ignored in the backlog.

## A late decision is still a decision

The cost of a slow decision is easy to underestimate, because nobody sends an invoice for it.

Do the arithmetic once. A decision that blocks three teams of eight people for three weeks costs 72 person-weeks. That is more than a year of work for one person, spent waiting, or worse, building on assumptions that will be thrown away. Multiply by the dozens of decisions a large program needs, and decision latency quietly becomes its biggest budget line.

And there is a hidden rule: not deciding is also a decision. It is the decision to keep spending while you wait. When nobody chooses, time chooses, usually the worst option, because the good ones expire first.

The human cost is just as real. Teams that wait too often stop proposing. The best people, the ones who like to move, leave first. What remains is a program that has learned to wait for permission, and calls it governance.

Slow decisions also feed another trap. Every choice you don't make keeps two options alive, and teams end up building both "just in case." That is how decision latency turns into the work in progress described in *The speed trap*.

## Early signals

You can hear this trap long before you can measure it:

- "Let's take it offline." Then nobody does.
- "We need to align with…" followed by a fourth committee.
- The same topic appears on three consecutive steering committee agendas.
- Teams build "both options, just in case."
- Nobody can name, in one sentence, who decides on a given topic.
- Meetings end with actions, but no decisions.
- People nod in the meeting and argue in the corridor afterward.
- The most heated debate of the month was about something reversible.
- A decision from last quarter is being "revisited" without any new fact.

## Before the car leaves the driveway

The answer isn't to decide alone and quickly about everything. It is to **make it obvious who decides, how fast, and what it would take to change our minds.**

1. **One name per decision.** Every significant decision has a single decider, named before the debate starts. Others are consulted, and some must agree, but one person says "we're going with this."
2. **Sort the doors.** Ask first: can we undo this cheaply? If yes, decide in days, at the lowest level that has the facts. Keep the slow, careful process for the handful of choices that are truly hard to reverse.
3. **Give every decision a "decide by" date.** Put the important ones on the program plan, next to the tasks they block. A decision past its date is escalated like any late milestone.
4. **Write it down in one page.** The context, the options, the choice, the reasons, and what would make us revisit it. Five minutes of writing saves five meetings of remembering.
5. **Reopen only with a new fact.** A decision log makes it simple: if nothing has changed since the decision, the decision stands. Disliking it is not a new fact.
6. **Disagree, then commit.** Everyone gets a real chance to argue. Once the decision is made, everyone executes it fully, including those who argued against it. That's what makes the debate safe.
7. **Make Abilene impossible.** Before a group decision, ask each person to write their view before anyone speaks. Ask the most senior person to speak last. Name someone to argue the opposite. Private doubts become visible while everyone is still on the porch.
8. **Remove gates, don't add them.** Merge reviews that check the same things. Give each gate a maximum waiting time. Push decisions down to the teams that have the information, and keep the steering committee for what only it can decide.
9. **Measure decision latency.** Track how long important decisions take, from question to answer. What gets measured gets shorter.

## What to avoid

- "Shared ownership" of a decision, which usually means no ownership.
- The same process for reversible and irreversible choices.
- Letting the most senior voice speak first.
- Unanimity as the goal: consent is enough.
- Decisions that live only in meeting memories.
- Reopening a decision because someone missed the meeting.
- Adding a new gate after every incident without removing an old one.

## The AI twist

AI makes decisions both easier and harder. Easier, because a comparison of options, a risk analysis or a summary of a long thread now takes minutes. Harder, because those same minutes produce more options, more analyses and more reasons to wait for one more. When studies become free, "let's study it further" becomes the most tempting way never to decide.

Meanwhile, AI speeds up everything around the decision. Code, tests and documents arrive faster than ever, which means decisions become the bottleneck even sooner. And coding agents can only act on what is written down: a decision that lives in someone's memory doesn't exist for them.

Used well, AI helps on the right side of the problem. It can prepare the one-page decision record, track decisions past their date, and flag a topic coming back to the agenda with no new fact.

## Saying "I'd rather not go to Abilene"

The best programs I've worked on were not run by people who never hesitated. They were run by people who made it clear, every time, who would decide and by when, and who made it safe to say, "Honestly, I don't think we should go."

So at your next steering committee, try one gentle question before the meeting closes: *what did we actually decide today, and who owns it?* If the room goes quiet, you may already be halfway to Abilene.

Have you ever been on a trip to Abilene, in a meeting room or anywhere else? I would genuinely love to hear your story.

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