How to Finish What You Start: A System for the Last 20 Percent

You know how to start. You have started plenty of things. The problem is the other end: the draft at 80 percent, the room that is painted except for the trim, the side project that works but was never shipped. If you want to know how to finish what you start, the honest first step is accepting that finishing is a different skill from starting, and that most productivity advice only teaches the first one.

Starting is a motivation problem. You solve it with momentum, smaller first steps, a timer, a body double. Finishing is a definition problem. You solve it by deciding what "done" actually means, and then making the last stretch small enough to survive a bad day.

Here is why the end is harder than the beginning, and a system that works on the pile you have right now.

Why the last 20 percent is the hardest part

The beginning of a project has a quality the end does not: it is vague. Vague is comfortable. You can work on "the website" for three hours and feel productive because any work counts as progress.

The last 20 percent is the opposite. It is specific, and specific work is exposing. The remaining items are usually the ones you have been quietly avoiding: the paragraph you cannot get right, the email to the person who might say no, the settings screen you have no design for. Those tasks did not end up last by accident. They ended up last because they are the hardest or the least pleasant, and you sorted them there without noticing.

There is a second reason. Early work has a fast feedback loop. Paint one wall and the room visibly changes. Fix the last three bugs and the app looks exactly the same as it did before, except now it is finished. The reward gets quieter precisely when the work gets harder, which is a terrible combination for anyone who runs on visible progress.

If this pattern is chronic rather than occasional, it is worth reading about why some brains abandon tasks at 80 percent, because the mechanism there is slightly different and the fixes are more specific.

Step 1: Define done before you need it

Most unfinished projects are not unfinished. They are undefined. Nobody ever wrote down what finished would look like, so the project cannot be completed, only abandoned.

"Done" has to be a sentence you could hand to someone else, and they would know whether you had achieved it. Compare these:

• Vague: get the spare room sorted
• Done: the boxes are gone, the desk is against the window, and I can walk from the door to the desk without stepping over anything

• Vague: finish the report
• Done: 2,000 words, three charts, sent to Priya

The second version of each is finishable. The first is a mood. You can spend six months in a mood.

Write the definition of done at the start, when you are enthusiastic and generous with yourself. If the project is already in flight, write it now, before you do another minute of work on it. This takes about ninety seconds and it is the single highest-leverage thing in this article.

One rule: the definition of done should describe the minimum acceptable version, not the best imaginable version. You can always improve something after it is finished. You cannot improve something that never got out.

Step 2: Build a finish list, not a to-do list

A to-do list is a list of everything. A finish list is different: it is the complete, specific set of things standing between this project and the definition of done you just wrote.

Sit down with the project open in front of you and write every remaining item. Not categories. Items. Not "polish the intro" but "rewrite the second paragraph so it does not repeat the first". Keep going until you genuinely cannot think of another one.

Two things happen when you do this.

First, the list is almost always shorter than the dread suggested. The thing you have been avoiding for five weeks turns out to be four items and about two hours. Dread scales with vagueness, not with effort, which is why the unopened project always feels bigger than the opened one.

Second, you find out what you were actually avoiding. It is usually one item. Everything else on the list is fine, and you were avoiding the whole project to avoid that single thing. Once it is written down in one line among nine other lines, it shrinks to its real size.

Step 3: Shrink the finish

Now make the remaining work small enough that a mediocre day cannot stop it.

Take the finish list and break anything longer than thirty minutes into pieces that are not. Thirty minutes is a useful ceiling because it fits in the gaps of a real day and because it is short enough that you will start it while unenthusiastic, which is the only state you can rely on.

Then schedule the first three pieces. Not the whole list. Three. Put them in actual slots on actual days, the way you would block time for anything else that matters. An item that lives only on a list competes with everything else you might do. An item in a slot competes with nothing.

If your estimate for the remaining work feels tight, double it. It is tight. Everyone's is, reliably and predictably, which is a well-documented bias rather than a personal failing.

Step 4: Handle the one thing you are avoiding

There is a specific item on your finish list that you skipped over twice while reading step two. That one.

Do it first, and do it badly on purpose.

Not "do it well quickly". Do it badly. Write the awkward email in three blunt sentences. Draft the paragraph you cannot get right as a placeholder that says roughly the correct thing. Build the settings screen ugly.

Bad-on-purpose works because the avoidance is almost never about effort. It is about the gap between the standard in your head and what you believe you can produce right now. Lower the standard explicitly, out loud, in advance, and the gap closes. You can raise it again on the next pass, and the next pass is much easier because now something exists to react to. Reacting to a bad version is a completely different task from generating a good one, and your brain treats it that way.

Step 5: Give the project a closing ritual

Projects do not end on their own. They trail off. Without a marker, the difference between "finished" and "abandoned in roughly finished condition" is invisible, and your brain files both of them the same way: as an open loop still quietly drawing power.

So make the ending explicit. Send it, publish it, hand it over, delete the working files, take the photo, close the tab group, cross it off somewhere you will see it. The specific ritual does not matter. What matters is that a moment exists which is unambiguously the end.

This is also what stops the endless-polish version of not-finishing, which is the same failure wearing a more respectable outfit. If done is defined and there is a closing ritual, the polish has a stopping point.

Step 6: Finish one before you start one

The reason you have several unfinished projects is not that you are bad at finishing. It is that starting feels better than finishing, and nothing in your system charges you for it.

So charge yourself. The rule: before a new project begins, one existing project reaches done or gets deliberately killed.

Deliberately killed is a real option and an underrated one. Some of your open projects are not stalled, they are over. You decided against them months ago and never said so. Saying so, out loud, is finishing. Write "not doing this, decided 25 August" and delete the files. That loop is now closed, and closed loops stop costing you attention.

This rule is the whole system in one line: if you only remember one thing about how to finish what you start, make it this one. Most people find that a third of their open projects can be killed on the spot, a third need one honest afternoon, and a third are genuinely worth finishing properly. The relief comes mostly from the first third.

What to do with the pile you already have

Everything above works on a project in flight. If you are looking at a backlog of eight half-finished things, do not run this system on all of them at once. Pick one. Preferably the one closest to done, because you need a win more than you need to be thorough.

Finish that one completely, including the closing ritual. Then decide about the rest with the finish-one-before-you-start-one rule doing the work. The backlog shrinks from both ends: you complete some, you kill some, and both count.

For the sorting itself, a three-pile triage of the whole backlog is faster than deciding project by project.

Frequently asked questions

Why do I lose interest right before the end? Because the reward gets quieter exactly when the work gets harder and more specific. Early work is vague and visibly changes things. Late work is precise and changes almost nothing visible. Nothing has gone wrong with you, the incentive curve is simply shaped badly, and defining done plus shrinking the last stretch is how you flatten it.

Is it better to finish a bad version or keep working until it is good? Finish the bad version. A finished thing can be improved, shown to someone, or learned from. An unfinished good thing is worth nothing to anyone including you. Almost every project that dies at 90 percent dies while being improved.

How many projects should I have open at once? Fewer than you currently do. One is ideal and usually unrealistic. Two is manageable. Beyond three, the projects start competing for the same attention and the ones you are not touching still cost you something, because an open loop does not go quiet just because you are ignoring it.

What if I genuinely do not know what done means for this project? That is a sign the project is really a direction, not a project. Directions cannot be finished, only pursued, so carve a finishable piece out of it. "Get fit" is a direction. "Run 5k without stopping" is a project. Give yourself the finishable version and the direction takes care of itself.

Does this work if I have already failed to finish this exact thing several times? Yes, and the repeated failures are useful information. Each abandoned attempt tells you where the wall is. If you have quit the same project three times at the same point, that point is the one item to do badly on purpose, and it was never the whole project stopping you.

The short version

Starting and finishing are different skills, and only one of them gets taught. To finish what you start: write down what done means in a sentence someone else could check, list every remaining item until the list stops feeling infinite, cut the pieces to thirty minutes, do the avoided one badly on purpose, mark the ending so your brain files it as closed, and do not start another until one is done or deliberately killed.

None of this requires more motivation. That is the point. Motivation is what you had at the beginning, and it is not coming back for the last 20 percent.