The Work

There is a pattern to a project that fails slowly. Everything is on track. Week after week, on track. Then, somewhere around week nine or ten, it becomes obvious to everybody at once that it is not going to land, and in the review afterwards, it turns out that three people had known since week three.

Nobody lied. They were answering the question they were asked, which was "are we on track?", and in week three, the honest answer to that question was yes.

What works

One meeting a week that ends in a decision.

Not a status update. The question is confidence rather than progress, and that difference is the whole chapter.

Alongside it: ideas getting tested rather than argued about, and an honest ending for every goal.

Do, test, repeat. Those three are one system, which is why they share a chapter, and we call it execution loops.

Confidence matters, progress follows

Progress is a fact about the past, and it is nearly always fine, because work has been done. Confidence is a judgement about the future, and it moves early.

Four questions, same order, every week.

One. For each goal, what is your confidence now, and what would increase it?

Two. So what are we deciding to do, learn via experiments, and repeat because it works?

Three. Issues list: what’s on it, and do we need help?

Four. Should this goal continue?

Then decide

Confidence is a reading. It is not a decision, and this is where most check-ins stop.

Every week, every goal gets one of: continue, change, or stop. Written down, with one line of why.

Two rules make it work.

1. Confidence falling for three weeks running forces a decision rather than another update.

2. Confidence is never used against the person who reported it.

Break the second one, and you will get the number you want to hear for the rest of the year.

One technical note that matters more than it should. On our template, confidence is a word in a column, not a colour in a cell. A red or amber background cannot be read by a person scanning back through a quarter, and it cannot be read at all by the AI in chapter ten.

The ending

Every goal gets ten minutes at the end. The end is often the end of the quarter, but could be sooner or later.

How it ended. Reached, stopped early, ran longer, superseded.

A grade. Excellent, good, fair, bad. Judged by a person, not computed. Nothing lands mysteriously on 0.7.

What it taught us, in the words a person writing the next goal here would want.

And anything you found that was real but did not serve this goal. This is the one that gets lost, every time. You went looking for one thing and noticed another, and because it was not what you were measuring, nobody wrote it down.

Write it as properties rather than a verdict. Not it failed, but what it actually did and under what conditions.

The adhesive held at about a tenth the strength we wanted, peeled off without residue, and could be reused four or five times.

A verdict answers one question, once. A set of properties answers questions nobody has asked yet.

Did the leading measure actually move the outcome? If not, the metrics model needs editing, not the team.

And a recommendation: stop, continue, or exploit. Exploit means it worked, and there is more where that came from, and it is the option almost everybody forgets. Most companies file a success and move on while the upside walks away.