How do you get anyone to use the optional step that makes the output better?

Almost nobody fills in an optional field. I knew that, I shipped one anyway, and now I'm watching the thing I added specifically to improve quality get skipped by most of the people it would help.

The setup, without naming anything: the tool takes one input and produces a result. Adding a second, closely related input makes the result measurably better in exactly the places people complain about. It costs the user nothing — no extra charge, same wait. It costs them one more step at the start.

I made it optional because the first run is where people quit. Asking for a second thing before anyone has seen a single result felt like the fastest possible way to lose them.

So now the version I'm actually proud of sits behind a field most people scroll past, and the version most people meet is the weaker one.

Four options, none of which I love:

• Make it required. Best output, worst first run.

• Ask for it on the second run, once they've seen a result and have a reason to care.

• Show a before/after of both outputs at the input step — which means explaining a feature to someone who hasn't used the product yet.

• Leave it alone and accept that most people see the weaker version.

I'm leaning toward the second-run ask, mainly because it's the only one where the user already understands what problem the extra step solves. The obvious hole in it is that I don't know whether people come back for a second run often enough for the timing to matter, and I'd rather not build the whole prompt just to find that out.

Has anyone here shipped an optional quality step and actually got adoption on it? I'm specifically trying to work out whether it was the timing that moved the number or the explanation.