I use a homebrewed spreadsheet which incorporates the Excel optimizer into the mix. Once I get results from that, I can then use that as guidance to my withdrawal calculator to make it practical. In other words, I think the Roth optimizer is great to help develop a plan but I wouldn't want to use it as both the Roth planner and the withdrawal calculator. Such a plan should be revisited from time to time.
As has been mentioned before, there are a lot of reasons to do Roth conversions. In my case, so far, what I've been optimizing on is a (remaining) lifetime increase in after-tax spending. Then I check the results for other scenarios such as one of us departing before the other and moving into single-filing category with a lower overall SS benefit, and a portion of what we will leave as an inheritance being tax free. And always, I compare against my default-no conversion plan.
One thing I've noticed, at least for our case, is that there are a lot of constraints I can place on the optimization that give nearly the same results. For example, as many have noted in this thread, constraining to the top of a bracket, constraining to below the first IRMAA level, max overall amount allowed to be converted in any given year, or completely unconstrained. Limited only by your imagination and the ability of the optimizer to actually find a solution. The main thing I see for the unconstrained case vs. any constraints I place on the process is the magnitude of the conversions themselves and how many years it takes to complete the conversion process. The unconstrained case, though, always gives the (slightly) highest overall after-tax spending capacity. And it does so in some surprising ways such as going into the first IRMAA bracket for a year or two at the beginning and even touching into a higher bracket than most rules-of-thumbs might have you do.
This year will be my first true conversion with me at age 63 and DW at 64. I underestimated my MAGI for 2024 for ACA, so for this first conversion, I'm merely converting up to the MAGI I originally estimated. It'll be plenty big enough I think no matter what other constraints I place for 2025 onwards. In 2025, we will still have a premium tax credits for the first few months of the year, but that will disappear when my DW starts SS in Q2. That will be mostly noise if I end up having to pay back the tax credits at the following tax season.
Finally, MaxiFI Planner has recently added a Roth Optimizer as an option to their tool. A good friend of mine has been a user of this tool for a number of years. After having watched one of their videos and actually sitting down with my friend as he ran the optimizer, I noticed something I hadn't considered. Today, even without any Roth conversions, we have a spending capacity that, so far, is quite a bit larger than we've been using. In the MaxiFI Planner examples, they break the spending into two parts: nondiscretionary and discretionary and allow Roth conversions to only increase the amount of discretionary spending. In 2025 as I consider the forward-going plan, I might consider a similar idea - it would require at least mentally breaking the portfolio into two parts: One part that provides nondiscretionary and one part that provides for discretionary spending and, hence is, subject to Roth conversions.
There are many ways to crack this nut. And based on what I've seen, Nth level optimization won't buy you much vs. having a plan of some sort in the first place.
Cheers.