The last part of a task often needs a decision about what is already good enough.

The final ten percent is rarely a technical problem. The work is mostly there, and what remains is a judgment nobody wants to make: whether this is finished, or whether one more pass would help. Because the judgment is uncomfortable, people substitute effort for it and keep polishing, which feels productive and postpones the decision. The ending arrives when you decide, not when the work runs out.

When you next stall near the end, stop working and name the standard instead. Ask what this needs to do for the person receiving it. If it does that, the remaining changes are for you rather than for them.