It Doesn't Just Simulate. It Searches for the Best Answer.

Two different questions, easy to conflate

"Will my plan work?" and "is my plan the best version of itself?" sound like the same question asked two ways. They aren't. The first is a grading question — you already have a plan, and you want to know its odds. The second is a search question — you're asking whether some other combination of choices, still within reach, would do better than the one you happened to land on.

A Monte Carlo simulation is built to answer the first question extremely well. It takes your plan exactly as it stands — your retirement age, your claiming strategy, your spending, your account balances — and runs it through many randomized sequences of market returns, reporting the percentage that hold up. That's a real, honest number about a real, specific plan. But it doesn't try anything else. It isn't asking whether retiring eight months later, or claiming Social Security two years earlier, or running a different Roth conversion schedule, would have produced a better outcome. It was never asked to.

A simulation grades the plan you already wrote. A search tries a bunch of different answers and tells you which one scored highest.

Why this takes a genuinely different kind of compute

The distinction isn't just semantic — it's a different computational shape entirely. A simulation holds the plan fixed and varies the market: one set of decisions, thousands of possible futures, a probability estimate as the output. A searched optimizer does close to the opposite. It holds a realistic range of futures constant (or accounts for it inside each candidate's own evaluation) and instead varies the decisions — running your plan over and over with a different claim age, a different conversion amount, a different spending level each time — and compares the resulting outcomes to find the one that wins.

That means a searched optimizer isn't a bigger or slower simulation. It's a different question, answered by trying many candidate plans instead of stress-testing one. Some of Deep Search's tools evaluate a full grid of combinations; others work backward from a target and narrow in on the specific lever value that hits it. Either way, the output isn't "here are the odds" — it's "here is the specific version of your plan that comes out ahead, and by how much."

What "searching" looks like in three of the actual tools

Deep Search — the Optimize tab's branded name, paid, distinct from the app's single-scenario simulation — is a rail of seven searched-optimization tools. Each one runs many what-if variants against your real plan and returns the best one it finds, not just the one you started with. A few concrete examples of what that search actually does:

• Couples SS optimizer. For a two-earner household, the claim-age decision isn't one choice — it's two, made together, and the interaction between them matters. This tool grid-searches all 81 combinations of user × spouse claim ages and returns the argmax — the single combination that actually comes out ahead. It won't crown a statistical tie as "the winner," either: a materiality floor holds it back from declaring victory when nothing beats your current claiming strategy by more than ordinary simulation noise, and it falls back to an honest "current is fine" instead.
• Solve-for-Goal. Most planning questions get asked backward from how they're usually answered. You don't actually want to know "what happens if I retire at 63" — you want to know "what's the latest age I can retire and still hit an 85% success rate." Solve-for-Goal is built for exactly that: pick a lever (retirement age, savings rate, monthly spending, whatever's adjustable) and a target metric and value, and it grid-searches and interpolates the lever value that actually hits the target — instead of you guessing a number, running it, and guessing again.
• Bridge Optimizer. The years between an early retirement and Medicare or full Social Security are usually the most tax-sensitive stretch of a plan — low-income years that are often the best window for Roth conversions, but also the years ACA subsidy exposure is most sensitive to extra taxable income. This tool runs a coordinated, multi-year search across both flat and year-shaped conversion ladders, with gain-harvesting on or off, priced against the actual subsidy math for that exposure — because the conversion schedule and the gain-harvesting decision aren't independent, and testing them one at a time would miss how they interact.

All seven tools share one search-and-report harness, so a "run" always narrates what it tried and shows the outcome it landed on being considered against the alternatives — not just a single number dropped on the page.

Why Deep Scan can't just add the other tools together

The natural instinct, once you've run a few of these tools individually, is to add up their gains: Couples SS found $40,000 more in legacy value, Bridge Optimizer found another $25,000, so running both should be worth $65,000. That arithmetic is wrong, and it's wrong for a specific, mechanical reason: the decisions these tools optimize aren't independent of each other.

A Roth conversion changes your taxable income in the years you're doing it. That changes which Social Security claim age is actually optimal, because claiming income and conversion income interact inside the same tax brackets and the same ACA subsidy cliff. A different claim age changes how many bridge years you have before Medicare and full benefits kick in, which changes what the "right" conversion schedule even looks like. Optimize each in isolation and you get two answers that were each correct on their own and can still be wrong together.

Deep Scan is the capstone tool built to handle exactly this. It runs the legacy-seeking searched optimizers together and reports the TRUE combined result through one final joint simulation with everything stacked — not a sum of the individual runs. That combined number is honestly, and usually, different from what you'd get by adding the parts. Sometimes it's higher than the naive sum, because one move clears a constraint that was quietly limiting another. Sometimes it's lower, because two moves were both drawing on the same limited resource — like the years available in the bridge — and the second one has less room left to work with than it did in isolation. Either way, it's the real number, not the flattering one.

What Deep Scan deliberately leaves out

Deep Scan runs the legacy-seeking optimizers together — but Spending Shape, the tool that searches for how much you can sustainably spend against a chosen success floor, is deliberately excluded from that combined run. Not an oversight — a design decision. Maximizing spending is a values trade: how much do you want to spend now against how much you want to leave behind, and there's no objectively "more optimal" answer to that question the way there is to a claim-age grid search. Folding Spending Shape into Deep Scan would mean the app quietly optimizing away a lifestyle choice on your behalf, treating "spend more" as a Pareto win the way a better claim age genuinely is. It isn't, and the tool doesn't pretend it is.

How this differs from Smart Moves

If you've read about Smart Moves, the distinction is worth being precise about, because the two features sound similar and do genuinely different work. Smart Moves scores a set of defined candidate moves — a catch-up contribution, a claiming-age shift, a staggered retirement between spouses — against your plan and ranks them by impact toward a goal you choose. It's a ranked list of options, each measured honestly, with the choice of which to act on left to you.

Deep Search runs full searches across a range of values for specific decisions and returns the actual best answer it found — not a list to weigh, but a number a grid search or interpolation landed on as the winner, subject to the same materiality-floor honesty that keeps it from calling a statistical tie a victory. Smart Moves tells you which levers are worth pulling. Deep Search tells you exactly how far to pull the ones that have a genuinely optimal setting.

What this doesn't do

Every number Deep Search produces still comes from a simulation of your plan against realistic market and tax assumptions — it's a projection, not a prediction, the same honest limit that applies to everything else this product produces. A searched optimizer can tell you which claim-age combination or conversion schedule performs best against the futures it tested; it can't guarantee tax law holds still, that markets behave the way historical data suggests, or that your own circumstances don't change the calculus next year. What it can do is replace "which of these should I try" with an actual answer, tested against your real numbers instead of a rule of thumb.

Common questions