Cash-flow planning vs portfolio modeling: what the difference actually means at the kitchen table.

A client sits down at the kitchen table. They are sixty-four. They have $2.1M in 401(k), $400K in a Roth, $180K in cash, a paid-off house, and one very specific question:

"If we retire next March, can we keep doing what we're doing?"

There are two ways to answer that question, and they do not produce the same answer. They do not even produce the same kind of answer. Most planning conversations conflate them, and the conflation is the reason a lot of plans feel both rigorous and useless.

This is the post about the difference.

The two paradigms, briefly

Portfolio modeling treats the plan as a math problem about asset growth. Inputs: starting balances, expected returns, volatility, contribution and withdrawal schedules, time horizon. Outputs: terminal wealth distributions, probability of success, asset-allocation glide paths. The frame is "will the portfolio survive?"

Cash-flow modeling treats the plan as a math problem about money in motion. Inputs: every income source by year (wages, Social Security, pension, RMDs, distributions, business income, rental, gain events), every outflow by year (essential, discretionary, healthcare, taxes, premiums), and the net effect on accounts year by year. Outputs: a year-by-year statement of where every dollar comes from, where it goes, and what's left in each account. The frame is "does the plan work?"

These look like the same question. They are not.

What portfolio modeling actually optimizes for

Portfolio modeling is built for a specific question: given an investment policy, what's the probability of meeting an objective. It's the right tool when the question really is about asset allocation, time horizon, and risk tolerance.

Where it earns its keep:

• Choosing a glide path for an accumulator who's twenty years from retirement
• Stress-testing the survival of a portfolio under specific volatility assumptions
• Sizing the buffer between essential and discretionary spend when the volatility matters more than the timing
• Comparing two policies — "60/40 vs 70/30" — on the basis of distributional outcomes

Where it stops being the right tool:

• Year-by-year decisions about which account to withdraw from
• Tax-aware sequencing of conversions, distributions, and gains
• Anything that involves a specific calendar event — a property sale, a child's college year, a planned partial retirement, a Medicare-eligibility year, an inheritance
• Anything where the answer depends on what happens to a specific tax line in a specific year

The portfolio model can absorb those things by collapsing them into an aggregate withdrawal stream. But the aggregate is where the information about timing dies, and the timing is most of what matters for clients close to or in retirement.

What cash-flow modeling actually optimizes for

Cash-flow modeling is built for a different question: given the plan, what does each year look like. It is the right tool when the question is structural — about timing, mix, sequence, and the interaction of variables that don't compose into a single rate.

Where it earns its keep:

• Withdrawal sequencing across taxable, tax-deferred, and tax-free accounts
• Roth conversion sizing in a year when the household is in a temporarily lower bracket
• Modeling the year a client claims Social Security, with the corresponding IRMAA bracket two years forward
• Pension-vs-lump-sum decisions where the answer depends on tax bracket in specific future years
• The "one big year" problem — a business sale, a property sale, a vested-equity event — and the years bracketing it
• Survivor scenarios where filing status and tax brackets change mid-plan

Where it stops being the right tool:

• Rough-cut feasibility for someone twenty years out, when most inputs are guesses anyway
• Conversations where the actual question is "what asset allocation matches my temperament," not "what does the plan look like year by year"

The cash-flow model is information-rich and assumption-heavy. It tells you a great deal about specific years, at the cost of requiring opinionated inputs about what each of those years contains.

Where they appear identical, and aren't

Both paradigms can produce a chart that looks like a portfolio balance over time. Both can run Monte Carlo. Both can spit out a "probability of success" number. The differences are in what those numbers actually mean.

A portfolio Monte Carlo of 10,000 trials, with a single composite withdrawal stream, will give you a probability that the portfolio survives the plan. It is mute on why the failed trials failed. It cannot tell you that 30% of the failed trials failed because of one specific kind of timing — say, the IRMAA bracket pushing the household into a tier where the surcharge ate the conversion savings — versus 70% that failed because of pure return-shortfall risk.

A cash-flow Monte Carlo with the same 10,000 trials, run against a year-by-year tax engine and an opinionated cash-flow structure, will produce the same probability of success. But it will also tell you which specific decisions in which specific years are eating the failure trials, and which decisions in which years are buying you the survival ones.

The probabilities can match. The information content does not.

The kitchen-table frame

I keep coming back to the kitchen-table frame because that's where these conversations actually happen. Across a table, with two people who want to know whether they can retire next March.

What they ask is not "what is my probability of meeting my goal." That is a question they have learned to phrase, sometimes, because we have taught them to phrase it. What they actually want to know is some combination of:

• "Can we keep doing what we're doing?"
• "What changes when one of us isn't here?"
• "What happens if the market drops 30% the year after we retire?"
• "What does the year my pension starts look like?"
• "How much can we give to the kids without it costing us later?"
• "What does Tom retiring two years before me actually do?"

These are cash-flow questions, all of them. They are about specific years, specific decisions, specific changes. The portfolio-model answer to most of them is some flavor of "your probability of success goes from 89% to 86%." That number is not wrong. It is not what the question is asking.

A cash-flow model answers them in their own terms. "Here is the year your pension starts. Here is what your tax bracket looks like that year. Here is the conversion we'd run that year, and here is what your IRMAA looks like in 2030 because of it. Here is the survivor year — the year after Tom is gone — and here is what changes in your tax bracket because of the filing-status flip."

The conversation is different because the language is different.

Withdrawal sequencing as the bridge

The clearest place to see the difference between the two paradigms is in withdrawal sequencing. The decision is: in any given year, which account do you draw from?

The portfolio-model answer tends to be a heuristic — "spend taxable first, then tax-deferred, then Roth" — sometimes refined into a tax-bracket-aware version of the same heuristic. It produces a defensible average-case behavior across many years and many clients.

The cash-flow answer is computed. In any given year, the question is: what's the marginal cost of the next dollar drawn from each account, including federal tax, state tax, Social Security taxability, IRMAA bracket two years forward, AMT (where it applies), QBI threshold (where it applies), and the long-term effect on the residual balance in each account. The right account to draw from is whichever produces the lowest total cost on that dollar in that year.

This is not a clever trick. It is just the computation. If you do not run it, you are using a heuristic. The heuristic is fine in many cases and wrong in some, and the wrong cases are usually the high-income years where the cost of being wrong is largest.

Monte Carlo on cash flow vs. on portfolio returns

A future spoke will go deep on this, but the headline is worth landing here:

Portfolio Monte Carlo varies investment returns. Cash-flow Monte Carlo varies investment returns and everything downstream of investment returns — tax outcomes, conversion sizes adjusted to brackets, RMD trajectories, IRMAA bracket placements. The first answers "will the portfolio survive." The second answers "will the plan survive, including all the decisions the plan implies."

The probability of success can come out the same. The trials that fail will not be the same trials. And the structural fixes you'd make in response — "don't run the conversion in this kind of year" — only become visible when the simulation knows about the conversion in the first place.

Why most planning tools blur this

The reason most planning conversations conflate the two paradigms is that the tools blur them. A typical planning tool will:

1. Take a portfolio model as the core engine
2. Layer a tax estimate on top, often as a separate calculation that runs after the portfolio model rather than alongside it
3. Approximate cash flow as a withdrawal stream from the portfolio model
4. Run Monte Carlo on the portfolio returns, with the tax estimate held constant per trial

Each of those steps is a reasonable simplification on its own. Together, they produce a tool that looks like it does both things and actually does one of them with cosmetic gestures toward the other.

The test is simple to state: in the simulation, when a Roth conversion lands in a given year, does the IRMAA premium two years forward update? Does the marginal tax rate on the next dollar of withdrawal in that year update? Does the AMT calculation for that year reflect the conversion? If any of those answers is no, the tool is doing portfolio modeling with tax cosmetics, not cash-flow modeling.

We built Foundry Planning so that the tax engine, the cash-flow engine, and the Monte Carlo engine are not three things stitched together — they're one engine. Every plan number is computed from the same source of truth on every keystroke, on every trial. That's the bar. It is not the bar most planning tools clear, and it took rebuilding the engine from the bottom to get there.

What this means for how you have the conversation

I run plans cash-flow-first because the conversations clients want to have are cash-flow conversations. The portfolio piece is part of it — the asset allocation has to make sense, the volatility has to be sized to the household's tolerance for it — but the portfolio is one input, not the frame.

The frame, at the kitchen table, is: here is your year. Here is the year after that. Here is what changes when the pension starts, when Social Security starts, when an RMD wave starts, when one of you isn't here anymore. Here is what we'd recommend for this year specifically, and here is the math.

If your current tool is shaped around portfolio modeling and you find yourself translating its outputs into cash-flow language by hand in the meeting, the tool is doing less work than it should be. The translation is the work. The tool should be doing it.

What's next in this cluster

This pillar opens onto the cash-flow spoke series. Future spokes will cover withdrawal sequencing in depth, RMD timing as the 73→75 transition lands, Social Security claiming as a cash-flow decision, the difference between Monte Carlo on portfolio returns and Monte Carlo on cash flow, and the "one big year" modeling pattern for business-owner clients.

In the tax cluster, the IRMAA spoke is the most concrete example of the cash-flow paradigm working: every decision against an IRMAA bracket is a cash-flow decision, computed across years, with the wrong answer hiding under the heuristic.

If you are choosing a planning tool — or thinking about whether the one you have is the right one for the kind of work you actually do — the tooling pillar walks through the evaluation framework I'd use.

If you'd like to see what cash-flow-first planning looks like in the room, we should talk.

---

This is the cash-flow cluster pillar. Working notes for future spokes are in the editorial backlog; if there's a specific cash-flow modeling question you'd like covered, hello@foundryplanning.com.