Ship at 70 Percent

The draft is done. It has been done for eleven days.

What's happened since is reading it again, moving a section, moving it back, and deciding to sit with it over the weekend. Nothing in the last eleven days made it meaningfully better. Everyone involved knows this, including you.

The stall is rarely about the work. It's about not yet feeling certain enough to be finished, and that feeling has a price nobody quotes.

The last thirty percent is the expensive part

Getting to roughly 70 percent confidence is usually fast. The information is available, the shape is clear, and the obvious problems have been dealt with.

Everything after that gets slow and stays slow. The remaining questions are the ones that resist desk research, because they depend on how something lands in the world. You can spend a week narrowing them and end up at 75 percent, which is not a week well spent.

Jeff Bezos framed the cost directly in his shareholder letters: most decisions should be made with around 70 percent of the information you wish you had. Waiting for 90 percent is slow, and being slow is itself a failure mode. It just isn't one that shows up in a post-mortem, because nobody writes a report about the thing that shipped three weeks late and fine.

Being wrong at 70 percent and correcting quickly beats being right at 90 percent and arriving late. That's true of decisions and it's true of work.

First, check which door you're standing in

The rule has a limit, and the limit is the useful part.

Bezos splits decisions into two kinds. Two-way doors can be walked back. Try it, and if it's wrong, reverse it at modest cost. One-way doors cannot: the money is spent, the relationship is over, the announcement is public.

One-way doors deserve the slow, careful, gather-more-information process. They're rare.

The trouble is that most people run reversible decisions through irreversible rigor. It feels responsible. It's actually a category error, and it's expensive precisely because two-way doors are the overwhelming majority of what you face. The pricing page, the draft, the new format, the first hire on a trial basis. All reversible, most of them treated as if they aren't.

So the question comes before the deliberation: can I undo this? If yes, decide now, at 70 percent, and put the effort into noticing quickly whether you were wrong. That's Ship at 70 Percent, and the skill that makes it safe is not better foresight. It's a tight feedback loop.

Which is more specific than paying attention. It means naming in advance what would tell you the call was wrong, and when you'll look. Ship the pricing change and check conversion at seven days, rather than whenever you happen to wonder about it. Without that, moving at 70 percent isn't speed. It's a decision nobody ever revisits, which is the version of this advice that does real damage.

Hold the opinion strongly, then attack it

There's an obvious objection: deciding at 70 percent sounds like deciding carelessly.

Paul Saffo's formulation is the answer to it. Strong opinions, weakly held. Commit to a confident first guess on the information you already have, state it plainly, and then go hunting for the thing that proves it wrong.

Both halves are load-bearing, and each fails on its own. A weak opinion weakly held is just fence-sitting with better manners. A strong opinion strongly held is how people defend a plan long after it stopped matching reality.

The loop is: forecast with conviction, look hard for the data point that doesn't fit, revise, run it again. A sequence of rough guesses, each genuinely tested, converges faster than one careful guess defended.

The hard part is being willing to be wrong out loud, in front of people who watched you be confident last week. That's Strong Opinions, Weakly Held, and the willingness is what keeps a plan current instead of defended.

What eleven days of not shipping actually bought

Nothing, but that's not the real cost.

The real cost is that everything the draft would have taught you is eleven days late. How people read it, which part gets quoted back, the objection you didn't anticipate. None of that was available at the desk. All of it was available within a day of shipping.

Polishing toward certainty and learning from contact are alternatives, not a sequence. Time spent on one is not spent on the other, and only one of them tells you something you didn't already know.

The rule, and where to break it

For anything you can reverse: decide at 70 percent, ship it, and watch closely. Reserve real polish for the few things that genuinely move the outcome, and let the rest go out at 70.

For anything you can't reverse: take the time. Gather the information. That's what the saved time is for.

Then go look at the draft that's been done for eleven days, and ask which door it is. It's a two-way door. It has been the whole time.