Spreadsheet
Withdrawal Order Sequencer
Same money, same investments, same spending. Which account you reach into first decides how much of it reaches you.
Download the worksheet
Opens in Microsoft Excel, Google Sheets, Apple Numbers, and LibreOffice Calc. To use it in Google Sheets, upload the file to Drive and open it with Sheets.
What the order is worth
$1,500,000 across three account types, spending $80,000 real a year
- Best order, lifetime tax
- $330,365
- Worst order, lifetime tax
- $488,087
- Difference
- $157,722
Taxable, then deferred, then Roth lasts 30 years. Roth first lasts 29. Nothing about the portfolio or the spending changed between those two rows.
Why the order matters at all
Three account types, three different tax treatments. A taxable brokerage account is taxed every year on what it distributes, whether or not you wanted the cash, and again on the gain when you sell. A tax-deferred account costs nothing until you withdraw, at which point every dollar is ordinary income at your normal rate. A Roth is never taxed again.
So one account has a running cost you cannot switch off, one has a cost you choose when to pay, and one has none. Spend down the one with the running cost, defer the one you control, and let the untaxed one compound for as long as possible. That is the entire argument, and this sheet is it projected forward year by year with every step visible.
What is inside
- Four projection tabs, one per ordering, 45 rows each with every intermediate column shown: the gross withdrawal, the gross-up for tax, the dividend drag on the taxable account, the cost basis amortizing as you sell, and the balances rolling forward.
- Compare, the four results side by side, ranked by years funded with lifetime tax breaking a tie. The ranking rule is written on the sheet rather than buried in a formula.
- Nothing hidden. If a number looks wrong you can follow it back to the row it came from, which is the point of a worksheet rather than a calculator.
The gross-up most people skip
To have a dollar in hand from a tax-deferred account you have to withdraw more than a dollar. At a 22% ordinary rate you withdraw about 1.28 dollars for every one you spend. Over decades that compounds into a materially faster depletion than a naive projection shows, and it is the single most common omission in retirement spreadsheets.
What it deliberately does not do
No tax brackets. You give it one ordinary rate and one capital gains rate and it applies them flat. This is a real limitation and it cuts against the sheet's own conclusion: the best strategy in practice is to spend the taxable account and take just enough from the tax-deferred account each year to fill the cheap brackets, and a flat-rate model cannot represent that refinement. Note that it is also not the proportional ordering, which draws by account size and so touches the Roth from year one.
It also omits required minimum distributions, Social Security, state tax, and the income thresholds that affect Medicare premiums. Each is real and each argues, at the margin, for drawing somewhat more from the tax-deferred account early than a strict ordering would suggest.
A warning about the default result
Every ordering runs out of money on the figures that ship with the file. That is not a bug in the model: it is a $1,500,000 portfolio spending $80,000 real a year for 45 years, which is a withdrawal rate the arithmetic does not support. The orderings are being compared on how long they last and what they hand over along the way. Fix the withdrawal rate first; the ordering is the second decision, not the first.
A correction this file produced
Building this worksheet meant checking its row arithmetic against the model behind the published essay, line by line. Three of the four orderings matched to the cent. The fourth did not, and the model turned out to be at fault: the proportional case was not amortizing cost basis as the taxable account was drawn down, which let the gain fraction collapse to zero and understated that strategy's lifetime tax by about $40,779 and credited it with a year of spending it had not earned. The model and the essay were corrected on October 7, 2026, and the corrected figures make the essay's conclusion stronger rather than weaker: the proportional order now comes second-worst of the four rather than running neck and neck with the best. The worksheet reproduces all four orderings to the cent.
Found an error, or think a calculation should work differently? Tell me. Corrections get published with a date attached, the same as they do for the research reports.