Investment AnalyzerDocs
Open Investment Analyzer

Projected success: what's a good number?

The Plan tab's headline is a percentage and a color. Reading them well takes three things: what kind of number it is, what a good one looks like for a household like yours, and what the runs that fail would feel like from the inside.

Where the percentage comes from is covered in Monte Carlo and scenarios, and what the model deliberately ignores in What the model leaves out. What follows builds on both.

One plan, many markets

Nobody knows the order of the next thirty years of returns, and — as the sequence-of-returns example on Monte Carlo and scenarios shows — the order is most of what decides whether a retirement works. A single projected return can't represent that. It gives one tidy answer to a question that doesn't have one.

So the plan doesn't make one forecast. It replays your plan against 1,000 randomly drawn market histories, month by month, from today to your horizon year. Some of those histories hand you a crash in your first retired decade; some hand you a boom; most are unremarkable.

Your plan is held fixed while they run. The same contributions go in, the same spending and goals come out, Social Security arrives on the same claiming dates. Only the market varies. That is what makes the number comparable between two edits: if it moved, it moved because you changed something.

Projected success is the share of those runs in which the portfolio was never emptied. Nothing more. A run that comes up even one dollar short is counted as failed and stays failed, even if a good decade afterwards would have refilled it — the running-out-is-permanent rule, is deliberate and which matters for everything below.

What's a good number?

It is closer to "probability of adjustment" than "probability you're fine"

Two things make this figure conservative by construction, and neither is an accident.

The first is that your spending never reacts to the market. Every one of the 1,000 runs spends the same real amount in the same year, whether the market has just boomed or just crashed — a household that would tighten its belt in a bad decade isn't modeled. That is the largest named simplification in What the model leaves out, and it understates success for anyone who could adjust their spending.

The second is the permanence rule above. A run where the portfolio runs a few thousand dollars short in a single year is recorded exactly like a run that emptied out in its first retired decade. Neither is given the chance to cut back and recover, which is what a real household would do.

Put together: a "failed" run is usually not a household that ended up destitute. It is a household that would have had to change something, and in this model was not allowed to. Derek Tharp, writing on Kitces.com in 2021, suggests reading the complement of the number that way — not a probability of failure but a probability of adjustment: "if we say there's a 50% probability of 'failure', it sounds potentially catastrophic; if we say there's a 50% probability of adjustment, it doesn't sound so bad."

That is the frame for everything that follows. The question isn't "will we be fine?" — it's "how likely is it that we have to change course, and how much would changing course cost us?"

Where the convention comes from

The plan grades the percentage in three colors, at two thresholds: green from 85%, bronze from 70%, red below 70%.

Published planning practice clusters a little above that. Morningstar's The State of Retirement Income: 2025 solves for a 90% probability of the money lasting a 30-year retirement, and reports that the highest starting withdrawal rate meeting that bar, for a household wanting a steady inflation-adjusted paycheck, is 3.9%. A 2021 Kitces.com article by James Yaworski describes typical targets as a band: "industry practices vary from advisor to advisor, utilized success rates tend to fall between 85% and 95%", and notes that "a 90% probability of success is often a popular choice".

Green here starts at the bottom of that band rather than the middle, on purpose. The band is where advisors set a bar for implementing a plan; this screen is something a household looks at repeatedly for decades, and a threshold that paints an ordinary plan bronze every time it is opened teaches nothing.

When 85%, or lower into bronze, is reasonable

Four situations, in rough order of how much weight they carry.

  • You are years from retiring and will re-run this. Today's number is a checkpoint, not a one-shot bet. A working household that re-checks annually gets many chances to adjust while adjusting is still cheap.
  • Guaranteed income covers a large share of the baseline. If Social Security and a pension cover most of what you spend, a shortfall is a trim rather than ruin, and the arithmetic bounds it. Tharp's example: a couple spending $6,000 a month with $5,500 a month of Social Security cannot be forced to cut more than 8.3% of their lifestyle, whatever the portfolio does, because that is all of their spending the portfolio is responsible for.
  • A real share of the budget is discretionary. Wants and Wishes goals, the travel years, the second car. If you would genuinely drop them in a bad decade, the model's rigid spending is charging you for flexibility you have.
  • You hold reserves the plan doesn't count. Home equity, the option to work part-time, a family situation that could change. None of these are in the document unless you entered them.

The striking evidence for not over-targeting is Tharp's 2021 study, which ran retirements planned at 95%, 70%, 50% and even 20% probability of success, with ongoing adjustment along the way. Median lifetime spending came out nearly identical across all four. What differed was what was left at the end. Against a stated legacy goal of $200,000, the 95% plan never came up short and averaged almost $600,000 of terminal wealth — roughly three times what was asked of it. The 20% plan fell short of that goal about 24% of the time, and, in Tharp's words, "still not by much". For a household willing to adjust, the target is mostly an income-versus-legacy trade-off — a higher target is bought by spending less while you are alive.

When to want 90% or more

  • Spending is nearly all non-discretionary, with little guaranteed income under it. The 8.3% bound above runs the other way when the portfolio is funding everything: there is no cheap trim available.
  • You won't or can't adjust — or the adjustment would land on somebody else. A surviving spouse or a dependent adult child inherits the plan's fragility without having chosen it.
  • You are at, or just past, the retirement date. Sequence risk peaks in the first retired decade and there is no paycheck left to absorb a bad one.
  • The decision is irreversible. Resigning, claiming early, a large gift. A cushion is worth more when there is no way back.
  • A legacy is a goal in its own right, rather than whatever happens to be left.
  • You are unsure of the assumptions. This is the one people skip. The percentage is sensitive to the expected returns and volatility on the Assumptions card — see Default assumptions — and a cushion buys robustness against the model being wrong, not only against the market being unkind.

Why not 100%

Chasing 99–100% means planning for the worst percentile of a model. The cost is paid in advance and in real life: years more work, or a lower standard of living for the whole retirement, to insure against paths you would have adapted to. And it is not free at the other end either — that is the finding above, where the 95% plan averaged about three times the legacy it had asked for. Money left unspent is a cost, just a quiet one.

There is also a modeling reason. Squeezing the last few points means optimizing against the tail of a market model that draws each month independently from a normal distribution. Those last points are the least trustworthy part of the whole exercise.

Below 70%

Red is where the answer changes. The mitigations below are each worth something on their own; down here they generally have to be used together, or one of the two big inputs — the retirement date, or the baseline spending level — has to move. That is a different conversation from "which lever do we pull", and the plan is a good place to have it: the fastest way to find out which of the two is doing the damage is to change one and watch. The plan will tell you what each one is worth in your case; nobody can tell you in advance.

In comparing plans the difference between two plans is more trustworthy than the level of either one, for the reasons set out in Monte Carlo and scenarios.

If the bad runs happen

Failing runs are overwhelmingly the ones with poor returns early in retirement. That means the warning arrives early — in the first bad years, while small changes can still get you back on track — rather than later in retirement, when you have fewer options.

Below are the moves households make in response, each with the nearest thing the app can do to price it. Before testing any of them, press Snapshot on the plan screen: that freezes the current plan and its score as a scenario chip, so you have something to compare against and can get back to it with one click.

1 — Defer or skip discretionary spending in a down year. Push the Wish-tier trip or the new car out by two years; drop one of them entirely. In the app: the goals section already prices the tiers for you — the Needs only · + Wants · + Wishes chips at the top of it are the plan's score funding each layer in turn. To test one goal, click its row to switch it off (the checkbox at the start of the row does the same, and its tooltip reads Include in the plan), or open the goal and change When does it land?. Both leave the figures in the document, so a click puts them back.

2 — Guardrails. The best-known version is Guyton and Klinger's decision rules (Journal of Financial Planning, March 2006): when the withdrawal rate drifts about 20% above where it started, cut spending about 10%; when it drifts about 20% below, raise it; and skip the inflation raise after a losing year. The payoff is large. Morningstar's 2025 report, whose base case is a 3.9% fixed real starting withdrawal at a 90% bar, puts the guardrails figure at 5.2% on a 40% equity / 60% bond portfolio — the price being a paycheck that moves. Derek Tharp and Justin Fitzpatrick's 2024 critique, on Kitces.com, is worth knowing too: applied mechanically, the classic rules can demand brutal cuts in a severe downturn (in their backtest of a 1965 retirement, spending falling to 54% below plan), which is exactly why a floor of guaranteed income underneath matters. In the app: this plan does not simulate guardrails at all — its spending never flexes, which is the main reason the percentage is conservative for a flexible household. The nearest test is a scenario with a lower Household spending today on the baseline-spending card: could we live on less if we had to, and what does that do to the number?

3 — Raise the floor. Delaying Social Security to the latest claiming age the plan offers, 70, buys a bigger, inflation-adjusted, market-proof check for life; a pension or an annuity does the same job. The higher the floor, the smaller any forced cut can be — the 8.3% bound again. In the app: the claiming-age buttons on the Social Security card (62 to 70, per adult, with strategy presets), and the Pensions & annuities section.

4 — Work a little longer, or part-time early on. Even modest earnings in the first retired years cut withdrawals exactly when sequence risk is worst, and they usually extend contributions at the same time. In the app: the Retires in year (the field is labelled with the age it works out to) and Saving, both on each adult's household card, or an entry in Other income with a start and an end year for a few years of consulting.

5 — Don't sell shares at the bottom. The real-world version is holding a few years of spending in cash and bonds to draw on during a drawdown, so the equity side is never the thing being sold cheap. In the app: the model rebalances by assumption and has no notion of a bucket strategy, so it cannot score this directly. The allocation it uses is the one you hold, not a plan field — the Assumptions card shows the weights read-only, and you change them by changing the portfolio (or, for a balance-tracked account, its Allocation mix on the account itself). What you can test on that card is how much the plan depends on the assumptions: what it looks like if the expected return and volatility turn out to be wrong.

6 — Re-run the plan every year. Today's number is not a verdict on the rest of your life, and the saved plan re-scores against your live accounts every time you open it. A plan that drops sharply after a bad market is the signal to reach for lever 1 or 2 while the correction is still a small one — the same shortfall met a decade later has to be met out of a smaller portfolio and fewer remaining years.

7 — Last-resort reserves. Downsizing, home equity, moving somewhere cheaper. The plan has no house in it, so none of this is counted unless you put it there. In the app: there is no field for a one-off sum, but an Other income row can stand in for one. Set its start and end year to the year the money arrives, enter the sum divided by twelve as the monthly amount, and give the row a label that says what it is ("House sale, one-off") so that anyone else reading the plan can tell.

None of these are recommendations. They are the moves available, and the point of the scenario chips is that you can price most of them against your own household instead of against an average one.

Sources