The three-planner study, in full

This is the detail behind the short version: how each plan was built, how it was entered in each tool, and every figure we compared. The point of the exercise was to test the math, so every plan uses the same stated assumptions in every tool, and wherever a tool’s own defaults would add something the plan doesn’t have (long-term care costs, its own Medicare estimate, saving of leftover salary), we set it to match.

How we ran it

• Tools. Boldin and ProjectionLab (Premium), entered by hand in October 2026, against our engine as published with this study.
• Assumptions, every tool. 6% a year on investments, 4.5% on cash, 2.5% inflation, birthdays in January.
• Dollars. Future (nominal) dollars, the figure each tool prints for that year. Balances are savings only; homes are left out.
• Steady returns. The year-by-year tables use a constant return so they compare the math, not the markets. The odds of success are each tool’s own simulation.
• Reading the tables. Each balance is the end of that calendar year. “vs ours” is the other tool’s figure against ours: positive means it is higher.

Plan 1 — Alex and Jordan

Married, North Carolina, born January 1967 and January 1969. $1.55 million saved today (401(k)s $850,000 and $300,000, Roth IRA $100,000, brokerage $250,000, cash $50,000), saving $44,500 a year until both retire in January 2032. Spending $8,000 a month in today’s dollars; health insurance $900 a month before Medicare, then $400 a month each. Social Security $3,000 and $1,200 a month at full retirement age, both claimed at 67. The plan runs to 2059 for Alex and 2064 for Jordan.

Income tax, federal and state

* ProjectionLab counts the Medicare income surcharge as tax; we and Boldin count it with healthcare. Odds of success on random returns: ours 81.7%, Boldin 80%. On history, every start year succeeds in both our back-test and ProjectionLab’s.

What agrees. The nest egg at retirement to within 1%; the first decade of retirement to within 0.5% (Boldin) and 1.5% (ProjectionLab); taxes within $850 a year of Boldin through 2059; the order money is drawn, when Social Security and required withdrawals start, and the survivor step-down.

What differs. We checked North Carolina’s tax for 2042 by hand: ours matches the state’s rules to the dollar, and Boldin’s lower figure comes from growing the state’s fixed $25,500 standard deduction with inflation. ProjectionLab’s later taxes run higher because it holds surplus required withdrawals as cash and taxes the interest; its balances run higher because it takes each year’s withdrawal later in the year. After the first death, Boldin keeps the deceased spouse’s 401(k) as a separate account, where we pool the household’s pre-tax savings.

Plan 2 — Sam

Single, Georgia, born January 1971. Retires at 58 (January 2029) and plans to 92. Saves $32,500 a year in a 401(k) on a $140,000 salary until then. Today: 401(k) $900,000, Roth IRA $150,000, brokerage $400,000, cash $60,000. A $2,000-a-month pension from 60 with no raises; Social Security $2,800 a month at full retirement age, claimed at 70. Spending $6,500 a month in today’s dollars. Marketplace health insurance at $1,100 a month before 65, with the premium credit; Medicare $400 a month from 65. Roth conversions of $40,000 a year (today’s dollars) from 65 to 69. A $450,000 home sold at 66 and replaced with a $300,000 one (today’s dollars). A $40,000 car at 62.

Boldin doesn’t estimate the marketplace credit, so its run was given our after-credit health costs; ProjectionLab priced the credit, the car and the home sale itself.

Taxes

Each column pair uses the same definition as the tool beside it. Odds of success on random returns: ours 87.2%, Boldin 89%.

What agrees. Boldin and our engine agree within 0.4% every year through the downsize. The home’s sale price ($640,252) is the same in all three tools; Social Security at 70 ($3,472 a month in today’s dollars), the Roth conversion amounts and the 2037 federal tax ($5,879) agree; the years either side of 65 agree on tax within $700 of Boldin.

What differs. From 62, Georgia excludes up to $35,000 of retirement income a year, and $65,000 from 65; Boldin taxes it in full (2041: $23,362 against our $14,943), which is most of the later gap between us. ProjectionLab estimates a larger marketplace credit than ours once Sam’s pension starts (its net premium in 2031: $1,540; ours about $3,000), and takes withdrawals later in the year; together those account for its lead. In 2029–30 Sam’s income is below the federal poverty line, and Georgia hasn’t expanded Medicaid, so there is no credit and no Medicaid; ProjectionLab and our engine both charge the full premium.

Plan 3 — Morgan and Riley

A married couple 12 years apart, born January 1963 and January 1975. Morgan retires at 64 (January 2027); Riley works to 60 (January 2035) on $180,000 and covers the household’s health insurance until then, then buys marketplace coverage at $1,000 a month to 65. Today: Morgan’s 401(k) $2,000,000 and Roth $200,000, Riley’s 401(k) $700,000 and Roth $100,000, joint brokerage $1,200,000, cash $150,000. Social Security $3,900 at full retirement age claimed at 70 (Morgan), $1,500 claimed at 67 (Riley, with the spousal top-up). Spending $12,000 a month in today’s dollars, $8,400 for the survivor. Medicare $400 a month each. California until Riley retires, then Nevada. The plan runs to 2051 for Morgan and 2070 for Riley.

ProjectionLab can’t change state partway through a plan, so its run stays in California and is compared with our engine kept in California:

Taxes once both are retired (California)

Odds of success on random returns: ours 98.7%, Boldin 99%.

What agrees. The 2038 required withdrawal: ours $148,050, Boldin $148,191 — both use the IRS joint-life table for a spouse more than ten years younger. Morgan’s Social Security at 70 ($4,836 a month in today’s dollars) and Riley’s spousal top-up, in all three tools. The final balance within 1% of Boldin; 2033–2042 within 1.5% of ProjectionLab.

What differs. The early gap is a timing choice: while Riley still works, our engine counts Riley’s future 401(k) saving in the household’s starting figure, where the other tools add each contribution as it’s made. While Riley works, the tools also treat working-year taxes differently (Boldin includes payroll tax). For the first Medicare surcharge years we assume the household reports the drop in income when Morgan stops working (Form SSA-44); Boldin prices them on the last working year. ProjectionLab treats 2051, Morgan’s life-expectancy year, as the first survivor year, and saves leftover income as taxed cash, which is why its late-plan taxes run higher.

The conventions that explain the gaps

Each is a reasonable choice that a careful tool could make either way. The two state-rule differences (Georgia’s exclusion and North Carolina’s deduction) are covered in the plans above. At the end of each plan, our engine finished lowest on Alex and Jordan and in between on Sam; on Morgan and Riley it finished just below Boldin and just above ProjectionLab.

What the study changed in our engine

Lining tools up year by year is the best way to find rough edges, and we found a few. Each was fixed within a day, and every figure on this page includes the fixes:

• Future tax tables now grow from January 1 of the tax year, not from today.
• A home sale and later one-time costs now count the months until they happen, not whole years.
• In states that didn’t expand Medicaid, income below the poverty line now gets no premium credit, as the law says.
• The marketplace credit now counts an older spouse’s required withdrawals.

Two conventions we kept on purpose: the 10% early-withdrawal penalty in the year you turn 59½ covers the whole year (it follows from taking the year’s withdrawal at its start), and the first two Medicare surcharge years assume you report the drop in income at retirement. The changelog has each change and who it affects.

Run it yourself

These are the exact plan files our figures come from. Open one in the app with Inputs → Restore from a file (it replaces the plan saved in your browser, so back yours up first with Back up everything), then expand any year in the Projection table and press Check this year yourself.

The files hold October 2026 balances. Opened in a later month, the app counts from that month, so its figures shift a little from the tables above.

What this doesn’t show

• Three plans, designed by us, run once in each tool in October 2026. The other tools change, and so does ours.
• Agreement isn’t proof that any tool is right; it’s evidence that independently built tools reach the same answers from the same inputs.
• Each tool’s figures were read from its own screens and reports; we don’t publish their screenshots. If we’ve recorded anything wrong, tell us and we’ll correct it.