I spent the last stretch building a tool that answers one question I get asked constantly in the FIRE world: how much of my traditional IRA should I convert to a Roth each year? Put another way: when does a Roth conversion make sense, and how much should you do? The tool runs a real optimization model, not a rule of thumb, and the interesting part is that once I started feeding it realistic numbers, the math kept repeating the same handful of lessons. Here they are, in plain English.
Quick disclaimer up front: I’m an operations-research guy, not your accountant. This is what the model showed me, not advice for your situation. Talk to a tax professional before you act on any of it.
1. The goal isn’t “max Roth.” It’s the smallest lifetime tax bill
It’s tempting to think the aim is to cram as much money into the Roth as possible. The model disagrees. Every dollar you convert gets taxed this year, so converting always costs you something now. You only come out ahead if that cost is smaller than the tax you’d otherwise pay later — when required minimum distributions force money out, or when whatever’s left in the traditional account eventually gets taxed.
So the right frame is: pay a little tax now, voluntarily and cheaply, to avoid a bigger forced tax bill later. The optimizer is constantly weighing those two against each other. If converting can’t beat the future bill, it tells you to leave the money alone.
2. The conversion window in your 60s is gold — fill the cheap brackets early
The single clearest pattern the model produces, over and over, is this shape: convert aggressively in the early retirement years, then coast.
Take a representative case — a 62-year-old who’s retired, sitting on a $1.2M traditional IRA, with Social Security not starting until 70. In those early years there’s almost no other taxable income, so the low tax brackets are sitting empty. The model fills them hard: it converts roughly $120,000 a year, deliberately running income up to the top of the 22% bracket (about $105,700 of taxable income in 2026 for a single filer). Then, once Social Security switches on at 70 and required distributions kick in, it drops down and just tops off the 12% bracket — about $50,400 of taxable income — for the rest of the plan.
That gap between when you retire and when Social Security and RMDs start is the most valuable real estate in the whole plan. Your brackets are never this empty again. Filling them while they’re cheap is the entire game.
3. RMDs are the villain the plan is built around
Required minimum distributions start at 73 or 75 depending on your birth year, and they’re a percentage of your traditional balance that the IRS forces you to withdraw and pay tax on, whether you need the money or not. The bigger your traditional account when they begin, the bigger those forced withdrawals — and they can shove you into higher brackets in your 70s and 80s exactly when you’ve lost the flexibility to do anything about it.
Conversions are the pressure valve. Every dollar you move to Roth in your 60s is a dollar that isn’t in the traditional account compounding into a larger forced distribution later. The model is essentially playing defense against your future RMDs, and the earlier it can act, the cheaper that defense is.
4. Social Security quietly eats your cheap brackets
This one surprised me even though it’s obvious in hindsight. Once Social Security starts, a big chunk of it is taxable, and it fills your low brackets before any conversion does. So the room you have left to convert cheaply shrinks the moment those checks begin.
It threw me when I first read the output: in one year the model recommended converting less than the required distribution that same year. My gut said that was a mistake. It wasn’t. The RMD is forced; the conversion is the optional top-up on top of it. By that age, Social Security plus the forced RMD had already filled most of the 12% bracket, leaving only a little headroom to convert before spilling into 22%. Different things, doing different jobs — and a good reminder that “convert” and “withdraw” are not the same line on the page.
5. Your assumptions drive the answer more than you’d like
A few inputs quietly do most of the work, and it’s worth knowing which ones:
- The future tax rate on what’s left. What you assume the un-converted balance will eventually be taxed at — by you later, or by your heirs — is the lever that decides how much converting is worth. Assume a high future rate and the model converts aggressively; assume a low one and it barely converts at all.
- Single vs. married brackets. Married-filing-jointly brackets are roughly twice as wide as single, so couples can convert far more per year before hitting the next rate. Worth knowing the brackets actually change, and by how much — the 2026 figures are here.
- How you value future tax. This is the sneaky one. Whether you treat a tax dollar in 2050 as just as painful as one today, or discount it, changed the recommended conversion by about 2.5x on the same inputs. A buried assumption, an enormous swing.
The takeaway isn’t “the tool is fragile.” It’s that anyone who hands you a single confident number for your Roth conversion without showing you these dials is overselling. The honest answer is a plan plus the assumptions it rests on.
6. The plan is dynamic — which is why a rule of thumb falls short
Most Roth conversion advice you’ll read is a static rule: “convert to the top of the 12% bracket every year,” or “convert a flat $40,000 annually until RMDs start.” Those aren’t terrible, but the model shows why they leave money on the table. The right conversion amount simply isn’t constant. It’s large in the empty early years, it drops the moment Social Security starts filling your brackets, and it drifts again as your required distributions grow and your balance shrinks. A fixed rule either under-converts during the golden pre-Social-Security window — the one time your brackets are wide open — or keeps mechanically converting into expensive brackets long after cheaper room has disappeared.
What an optimizer does that a rule can’t is solve the entire timeline at once. Each year’s recommended conversion already accounts for every other year — the Social Security that starts at 70, the RMDs that begin at 73 or 75, the balance that’s still compounding the whole time. That’s the real difference between a rule and a plan: a rule looks at this year, a plan looks at all of them together and works backward.
There’s a married wrinkle worth flagging too. Because joint brackets are roughly twice as wide, couples have far more cheap room to fill, so their optimal plans tend to front-load even harder. And there’s a sharper, sadder version of the same point that’s worth keeping in your own head even though this first version of the tool doesn’t model it: when one spouse passes, the survivor usually files as single, with brackets half as wide, on a similar pile of income. That “widow’s bracket cliff” is one of the strongest real-world arguments for converting earlier while you’re still filing jointly. The model uses a single filing status throughout for now, but the instinct it keeps confirming — convert while your brackets are wide — points the same direction.
7. What the model does not do (yet)
Being straight about the limits: this first version sizes your conversions and reports the tax. It does not yet model your actual living expenses, or which account you pay the conversion tax from. That last point matters a lot in real life — paying the tax out of a taxable brokerage account, rather than out of the converted money itself, is a big part of what makes conversions powerful. That’s the next layer of work, and I’d rather ship a smaller tool that’s honest than a bigger one that hand-waves.
It’s also federal tax only, uses a single assumed return, and doesn’t touch IRMAA or the early-withdrawal penalty. Plenty of room to grow. For the official rules behind any of this, the IRS pages on Roth IRAs and required minimum distributions are the source of truth.
Where to go next
If you want the bigger-picture setup, start with the Roth conversion ladder and its blind spot and then how much to convert each year. If you’re the type who wants to see the actual optimization under the hood, I wrote up the modeling lessons from building it too. And the free calculator itself is coming soon — this whole series is the work leading up to it.
