What you actually need to retire — and how likely it is to hold

How many months, and why it depends on your job

Three months, six months and nine months are not arbitrary. The right size for an emergency fund is however long your income realistically takes to come back, and that varies enormously by how you earn.

A government or PSU salary is the most secure income in the country. Redundancy is close to unheard of, salaries are paid on time, and the risk you are covering is a medical event or a family emergency rather than unemployment. Three months of spending is enough.

A salaried private job comes with a notice period and then a job hunt. Two months of notice plus three or four months of searching is an ordinary experience, not a catastrophic one, and six months is what covers it.

A business or freelance income can simply go quiet. A client leaves, a project slips two quarters, a season is bad. Nothing has gone wrong and yet nothing has come in, and nine months is the figure that lets you sit through it without dismantling your investments.

One thing almost every calculator gets wrong here: the target has to cover your home loan EMI as well as your household spending. Household spending is compressible — you can stop eating out, defer a holiday, drop a subscription. An instalment is not. It is the bill that does not care that your income stopped, and leaving it out of an emergency fund understates the target by a third or more for most borrowers. That is why the EMI field on this page feeds the emergency fund target directly.

Why the house you live in is not part of your corpus

This is the most expensive mistake in Indian retirement planning, and it is an easy one to make, because a home genuinely is the largest asset most households own.

But a retirement corpus has exactly one job: to be converted into spending, gradually, for thirty or forty years. Your primary residence cannot do that job. You cannot sell a quarter of it in year three. You cannot sell all of it and carry on living in it. If you did sell, you would have to buy or rent somewhere else, which consumes most of what you released. The house is worth what it is worth — it is simply not spendable, and a corpus made partly of unspendable assets will run out earlier than the arithmetic says.

Most calculators handle this by having no field for it. That does not solve the problem; it causes it. Faced with a box marked “other assets” and a home worth a crore, people type the crore. This planner gives the residence its own clearly-marked field, records it, adds it to your net worth, and then keeps it out of every projection on the page. There is a test in our suite asserting that setting the residence to five crore leaves the Freedom Number, the projection and the required SIP byte-for-byte identical.

Property you rent out is a different asset entirely, because it produces income you can actually spend. Enter that rent under Assumptions and it reduces what the corpus has to cover.

What this planner does that the others do not

• One simulation, not two methods. Your goals, school fees and college funds are spent in the years they fall due, inside the same year loop that grows the money. The Freedom Number is then solved by bisection as “the corpus at retirement that makes this path survive”. Calculators that sum a present value for the target and then run a separate projection produce two figures that cannot both be right.
• NPS is locked until 60. Not a preference — a rule. And at 60, at least 40% has to buy an annuity that pays a level amount for life while inflation halves its purchasing power. All three of those facts are modelled.
• The home loan is amortised, not averaged. Month by month, through the same engine as our home loan EMI calculator, so the closure month here and the closure month there cannot disagree.
• Capital gains tax on real gains. Cost basis is tracked year by year and withdrawals are grossed up, so the money that lands is the money the year needs.
• A Monte Carlo that reports a probability. Returns are drawn per year, cash sleeves swing less than equity sleeves, and the seed travels in the shared link so the result is reproducible. If a simulation returns the same figure as the deterministic model, it is decoration.
• Stress tests with numbers in them. Each scenario is re-solved, so you get the extra corpus and the extra monthly saving it would take — not a row of identical failures.

Where these numbers come from

Every rate this planner assumes, and the published figure it was set against. Where the setting differs from the source, the last column says why — a citation that does not explain the gap is decoration. All of them are yours to change, and the number above will move when you do.

What this model does not do

These are the limits. They are listed because a model whose limits are hidden is asking to be trusted further than it deserves, and because every one of them is a place where your own judgement should override the arithmetic.

• Returns have no fat tails. Each year’s return is drawn from a lognormal, which is well behaved. Real markets crash harder and more often than that, and the same shock hits every sleeve in the same year here rather than rippling through them. Published work puts the resulting understatement of failure rates at 10–17 percentage points, so read the probability as an upper bound rather than a promise.
• Nothing mean-reverts. A bad decade is not followed by a good one, and a good one is not followed by a bad one. Each year is drawn independently.
• Spending is flat in real terms. Blanchett finds real retiree spending falls roughly 1% a year before rising again late in life. Modelling that would lower the number, so leaving it out errs cautious — but it is an assumption, not a finding.
• Tax is capital gains only. There is no slab arithmetic anywhere here. Debt-fund gains on units bought after April 2023 are taxed at your slab rate, which at these spending levels is usually well above 12.5%, so the tax charged is too low for a debt-heavy corpus. Pension and rental income are entered after tax for the same reason.
• Every withdrawal is treated as long-term. No holding-period test and no short-term rate.
• The three strategies differ in their return assumptions, not only in their shape. Bucket keeps an equity sleeve earning the working-years return right through retirement; Deep de-risks the whole corpus. That is the point of the comparison, but it means the rows are not the same money held three ways.
• Your EPF contribution grows at the same rate as your SIP step-up. There is no separate salary-growth input.
• One rate covers both EPF and PPF, and the two are 1.15 points apart.
• This is arithmetic, not advice. It knows nothing about your job security, your family, your health, or what you actually want the money for.

The uncomfortable part

Run the default example and the probability sits well below what the single headline number implies. That is not a bug, and it is not pessimism. A Freedom Number computed on average returns is, roughly, the corpus that works if every year is average — and because investment returns compound multiplicatively, the median outcome sits below the average one. Any calculator that shows you a deterministic number and a 90% success rate on the same assumptions is showing you one of the two incorrectly.

What actually improves the odds, in order: retiring later, spending less in retirement, saving more, and only then earning a higher return. The first two move the number far more than the fourth, which is the opposite of where most people put their attention.