Money in, money out

The plan is spending-driven. You never tell it a target income. You tell it what your household spends and what money arrives, and the portfolio funds whatever is left over — grossed up for the tax on the withdrawal.

That's the whole shape, and every field in the workbench is one side of it:

  money in     = what you save + Social Security + pensions + other income
  money out    = baseline spending + goals + health insurance
  the gap      = money in − money out
                 positive, and it goes into the portfolio
                 negative, and the portfolio pays it, plus the tax

This page walks each lever, in that order. For how the result is computed and read, see Monte Carlo and scenarios; for where a withdrawn dollar comes from, Taxes in the plan.

Money in

Your working years are one number

The plan does not model your salary, and it does not model what you spend while you're working. It models one quantity: net money going into the portfolio.

That isn't a gap. Salary minus living costs is what you save, so asking for all three would be asking the same question twice and inviting the answers to disagree. What you type is the answer: what you save each month, in today's dollars, on each adult's card in the Household section.

It's per adult, and it stops at that adult's own year they stop working. That one decision is what lets the plan describe the ordinary two-earner case — one of you retiring a few years before the other — without a single extra field. Saving becomes a step function, and the Household section reads it back to you above the cards:

Saving $4,500/mo → $1,800/mo in 2038 → $0 in 2041

There's no household "retirement year" to type anywhere. The household retires when the last adult does, and the card shows that year marked (derived).

Which of you moves matters. Move the last adult's date and everything keyed to the household moves with it in the same keystroke: their saving step, the derived household year, when baseline spending starts, the Roth conversion window, and any coverage set to start when the household retires. Move an earlier retiree's date and you move their own saving step and their own coverage — the household's year is a maximum, so it doesn't budge until the last adult's does.

Two consequences follow from modeling the working years this way, and both are worth knowing before they surprise you:

Money arriving before you retire is assumed to be saved in full. A rental or some consulting income entered as an income stream doesn't reduce your living costs — the plan doesn't have your living costs — so it lands on top of what you told it you save.
A working year whose goals outrun your saving draws on the portfolio. A 2030 tuition bill bigger than that year's saving plus income takes the difference out of investments, taxed like any other withdrawal. Your saving carries on; the shortfall comes out on top of it.

Both figures live in the Social Security section, one card per adult:

{Name}'s full-retirement-age benefit — the monthly amount at full retirement age, in today's dollars. This is the "primary insurance amount" your Social Security statement quotes. A child never has one; the plan won't let you give them one.
Claim age — a row of buttons from 62 to 70.

The claim age scales that benefit for life, by the standard schedule, and the card reads the consequence back beside the name — for someone claiming at 67, $3,120/mo in today's dollars · $861k collected by 90. The Household card doesn't repeat the inputs; it carries one derived sentence, "Social Security: claims at 67 for $3,120/mo — tune it in the Social Security section." The schedule itself:

Claim at	62	63	64	65	66	67	68	69	70
Benefit	70%	75%	80%	86.7%	93.3%	100%	108%	116%	124%

Claiming at 62 pays 70% of your figure forever; waiting to 70 pays 124%. The plan treats 67 as full retirement age for everyone. That's correct for anyone born in 1960 or later; for people born before that, whose real full retirement age falls a few months to a year earlier, the schedule is slightly off.

For a two-adult household there are four strategy presets that set both claim ages at once — both at full retirement age, both early at 62, the higher earner delaying to 70 while the other claims at 62, or both delaying to 70. They exist so you can try the four shapes of this decision in four clicks rather than sixteen. A one-adult household doesn't get them; there's nothing for a preset to coordinate.

An In the plan checkbox in the section's top right covers the whole program at once. Untick it and the plan collects no Social Security for anyone, in any year, while every benefit and claim age stays exactly where it was. It's the fastest way to ask what your plan looks like if this isn't there — and it's a different question from cutting the benefits, which you'd do by scaling each adult's figure.

One number that will not reconcile if you check it by hand: in the full tax mode, benefits arrive 8% lighter than benefit × claim factor, because that is where the plan charges the tax on them — Taxes in the plan says why it's charged there and not on the withdrawal.

Pensions and annuities

Any number of them. Each is a monthly check that starts when one person in your household reaches a given age — so the plan works the calendar year out from that person's birth year rather than asking you for it.

The field that matters most is the cost-of-living adjustment. Without one, the check never shrinks in dollars — but everything else gets more expensive, so it buys less every year. The plan says how much less, in the year it happens:

No COLA: by 2061 this check buys what $671 buys today.

That's a $1,100 pension starting in 2041 at 2.5% inflation, 20 years in — and 20 years is roughly how long a pension started in someone's early sixties actually pays, long enough for ordinary inflation to have done something visible. A pension with a COLA holds its value and shows no such line.

The survivor option (50% or 100%) is recorded and displayed, and it is not simulated. The plan has no mortality model — nobody dies mid-plan — so the election never fires and choosing 100% here will not change your odds. It's captured because it's a decision you make once and can't revisit, and because the section shows it. What a real 100% election costs you is a permanently smaller monthly check, which you'd model by entering the smaller amount.

Other income

A rental, part-time consulting, a trust distribution: anything with a start and an end. Each has a label, a monthly amount in today's dollars, a start year, and an optional end year — leave the end blank and it runs to the end of the plan.

These are flat in real terms: they keep their buying power. A stream that doesn't — a fixed annuity with no adjustment — is a pension without a COLA, and belongs in the section above where the erosion is modeled.

A pension card carries an In the plan checkbox; an income row has a bare one at the start of the row (Fund to a screen reader). Either way — as with goals and health insurance — you compare a plan with and without by unticking, rather than deleting the figures and retyping them later.

Money out

Baseline spending

One number: what your household spends per month, in today's dollars. Nothing feeds it automatically, and that's deliberate — what you'll spend in retirement is a judgment call, not an average of last year's transactions.

It's charged from the household's retirement year onwards, not before. (Before then, remember, your spending is already accounted for: it's inside the difference between your salary and your saving.)

Two things bend it.

The step-down as children leave

Each child in the plan has an age you mark them independent at. Once they pass it, baseline spending steps down by a percentage you set — 8% per child by default. Two children at 8% each means spending settles at 84% of your baseline once both have gone.

Combined, the step-downs are floored at 40% of baseline: a large family can't step its way down to nothing.

The retirement spending smile

This is the second bend, and it's the one people are most often surprised by, so it's worth some room.

What the research observes. Retirees' spending, measured in real terms, tends to decline through retirement rather than keeping pace with inflation. Spending is highest in the first years — travel, the projects that were waiting, the second home — then falls as people slow down, and falls again in the final phase, when discretionary spending drops away faster than health costs rise. The term "retirement spending smile" comes from David Blanchett's research on retirement consumption, which is the work most often cited for this pattern; the three phases are widely nicknamed go-go, slow-go and no-go. It's the "smile" because plotting total spending including health costs against age gives a curve that dips and then turns up at the end — though for most households the turn never quite reaches the starting level. The app's own three phases never turn back up, because health costs are a separate term here rather than part of the baseline.

What the app does with it. The smile is a multiplier on your baseline, keyed to your primary member's age, in three phases:

Phase	Age	Percent of baseline
Go-go	to 75	100%
Slow-go	75–85	85%
No-go	85+	78%

The percentages are yours to argue with — all three are editable, and the section reads the result back as 100% to 75 · 85% to 85 · 78% after. The phase ages are fixed at 75 and 85. That asymmetry is on purpose: how much your spending falls is a claim about your household, and the shape is a claim about the research. If you disagree with the shape entirely, one checkbox turns the smile off and holds spending flat in real terms for the whole plan — the more demanding of the two assumptions, since it never lets spending fall with age. The per-child step-down still applies either way; it isn't part of the smile.

Why health insurance is not in here. Because it's the opposite shape. Baseline spending tapers to 85% at 75 and 78% at 85; health costs rise with age. Folding a premium into your baseline would taper it exactly where it should be growing, and would keep charging it in the Medicare years where the figure changes. So it's a term of its own — see below.

Goals

A goal is a lump sum you want the plan to carry: a kitchen, four years of tuition, a wedding, a new car every eight years. Each has a label, an amount in today's dollars, a tier, and a timing shape.

Three timing shapes, and the form gives you exactly one control with three modes so a goal can't accidentally be two of them at once:

A single year — the amount, once, in that year.
A run of years — the amount in each year from the start to the end inclusive. Four years of tuition at $40,000 is $160,000, not $40,000.
Every N years — the amount in the start year and every N years after it, to the end of the plan.

Three tiers — need, want, wish. They cost the plan exactly the same; what they buy you is the ability to ask what the plan looks like funding only the needs, or only needs and wants, which the section shows as running totals per tier.

Goals fire in every year they land in, including years before you retire. A 2030 college bill is real whether or not anyone has stopped working, and the plan doesn't pretend otherwise.

Each row carries the checkbox, the label, the amount — editable in place, so you can watch the badge move while you argue about it — the cost badge, and a pencil that opens the full form. A newly added goal lands switched off, so adding one changes no number until you say so.

Parking versus deleting. Unticking a goal keeps it in the plan with all its figures and stops it being funded; the badge beside it goes on telling you what it would cost. Remove, in the edit form's footer behind a confirm, deletes it outright. Untick to ask what if we didn't? — it's reversible in one click. Remove when the figures were never right.

Health insurance

If you stop working before 65, you buy your own health cover until Medicare starts. For an early retirement this is routinely the largest single line in the budget, and it's the line that moves most when a retirement year moves — which is exactly the knob the workbench exists to turn. So it's modeled per adult, with its own section, rather than folded into your spending.

Each adult's card takes three things and derives the rest:

Before Medicare — the monthly premium plus expected out-of-pocket, in today's dollars. An unsubsidized marketplace plan for someone in their late fifties or early sixties is routinely four figures a month, per person.
From 65 — Part B, Part D and Medigap once Medicare starts, plus IRMAA if you expect to pay it.
Starts paying — when you retire, or when the household retires.

No dates are typed. Coverage starts on the retirement year that choice picks out, and switches to the Medicare figure the year that person turns 65. That's the whole reason this is its own section rather than a goal: a goal is pinned to calendar years you type, so moving a retirement year would leave a hand-built health insurance goal sitting where it was. The card says the answer back to you in a sentence:

Don pays for coverage 2038 → 2044 (7 years, ≈$116k in today's dollars), then Medicare from 2045

The two coverage choices differ only in a staggered household, and the difference is the common case. The first person to retire often stays on the working spouse's employer plan, and their bill starts when that person stops. Choosing when the household retires says so — and it carries past 65: such an adult is charged nothing at all in those years, not even the Medicare figure, and their Medicare figure begins in the household's retirement year rather than on their 65th birthday. That's what "the household's plan still covers them" means. If that isn't your arrangement, use when you retire and adjust the amount. (With three or more adults in the plan, "the household" means the last of all of them to stop, which may not be the person you had in mind.)

Three states, not two. Leave an adult on Not modeled and the plan says nothing about their coverage — which is different from entering $0, and that would assert their coverage is free. Untick In the plan on a card and the plan stops charging it while the card goes on telling you what it would cost. The × takes the person back to Not modeled entirely.

Medical inflation, above CPI sits below everyone's cards because it applies to all of them. It is excess inflation — points per year above your general inflation rate — and it compounds the health insurance figures and nothing else in the plan. Everything else in the model is flat in today's dollars, so 0 here prices the bridge flat like every other line. The default is 1.5%: health costs have run one to two and a half points a year faster than everything else for decades.

The compounding runs from today, not from the year the bridge starts, so how much it matters depends more on how far off your retirement is than on how long the bridge lasts. At the default rate, a seven-year bridge starting a decade from now costs about a quarter more than the same premiums priced flat; the same seven-year bridge starting next year costs only a few percent more.

Three related costs are deliberately out of scope and named rather than silently approximated: ACA premium subsidies and IRMAA, both of which depend on your income for the year, which depends on which account a withdrawal comes from — enter the amounts you expect to actually pay. And long-term care, which is a different kind of risk: a possible large cost late in life, not a monthly premium. Carry that as a goal if you want the plan to hold room for it. What the model leaves out is the full list, with the direction each simplification pushes your result.

Where each lever shows up

Every one of these lands in the Year by year table under the chart — Year, Ages, Income, Saving, Spending, Health, Goals, Taxes, Net flow, Balance, one row per calendar year. If a year looks wrong on the chart, that's the row that says why, and the Taxes cell opens to show what the year's money movement actually cost.