Thirty investors. Thirty different days of the month. Identical SIP — ₹10,000 every month into Nifty 50. After ten years, their final corpus differs by less than two months of one investor's SIP. After twenty years, the entire spread between best day and worst is one tenth of one percent of the corpus. After thirty years, the spread is eight basis points of XIRR. The day-of-month decision is a non-variable: it borrows the appearance of choice from a number that does not move the needle. This paper walks through the empirical proof, then explains why mechanism beats timing at every horizon — and what the investor should be optimising instead.
Most prospective SIP investors agonise over the start day. First of the month? Fifth? Tenth? Fifteenth? After the salary credits? Before the rent debit? The forums fill with theories: "buy when the market is low," "first-of-month is best because everyone else does it," "the seventh always lands near the bottom of the daily range."
Each theory has the surface texture of analysis. Each is functionally astrology. The data — decades of actual Nifty 50 closes, every trading day since 1996 — gives the same answer regardless of which day you pick: the path lands within a few basis points of every other path. The difference between the best day and the worst day, over twenty years of monthly investing, is approximately one-tenth of one percent of the final corpus.
The day-of-month decision is what statisticians call a non-variable: a parameter that appears in the question but does not appear in the answer. The compounding engine is too large for it. Twenty years of market path swamps any twenty-eight-day micro-pattern, and the law of averages closes whatever residual gap remains.
The cleanest way to test the proposition is to actually run it. Pick thirty hypothetical investors. Each one starts a SIP of ₹10,000 per month into Nifty 50, on a different calendar day — the first runs on day 1, the second on day 2, all the way to day 30. Same fund. Same amount. Same horizon. Different day. Run them through real Nifty 50 close prices, with actual transaction-date XIRR. Three horizons: 10, 20, and 30 years.
| Horizon | Best Day Corpus | Worst Day Corpus | Best−Worst Spread | XIRR Range |
|---|---|---|---|---|
| 10 years (May 2016 – Apr 2026) | ₹19.4L | ₹19.2L | ~₹14k | 10.78–10.92% |
| 20 years (May 2006 – Apr 2026) | ₹77.4L | ₹76.8L | ~₹53k | 10.54–10.62% |
| 30 years (Apr 1996 – Apr 2026) | ₹3.12 Cr | ₹3.09 Cr | ~₹3 L | 11.99–12.06% |
All three windows end Apr 30, 2026. Each row aggregates 30 SIP scenarios across 30 different start days. The best-vs-worst spread is the entire range across all 30 day choices. XIRR is annualised, day-count XIRR (Excel methodology) where available; calibrated monthly-annuity for windows that pre-date our daily-data feed. Source: Nifty 50 historical close prices.
A counter-intuitive but mathematical pattern: longer horizons have tighter XIRR spreads, not wider ones. The reason is simple — with more SIPs, the within-month price variation averages out more completely. By 30 years (361 monthly SIPs), the residual day-of-month effect has been smoothed almost to zero. The spread is mostly arithmetic noise from the precise day a single payment lands relative to the others.
A 20-year SIP investor who picks the empirically worst day-of-month forfeits, over their entire investing lifetime, less than the value of a single month's contribution. This is what we mean by non-variable: the parameter cannot move the outcome by an amount that matters. The interactive version of this experiment is at the SIP Timing Paradox calculator →, which lets you adjust the SIP amount and pick your own day.
Within any given month, Nifty 50 typically moves in a band of two to four percent from peak to trough. The day you SIP determines where in that band your purchase lands. So in a single month, the best-day buyer might pay 2% less per unit than the worst-day buyer — a real, observable advantage.
Then the second month happens. And the within-month band has shifted somewhere else. The day that was best in month one is not necessarily best in month two — the relationship is essentially random across months. A 2% advantage in one month, averaged with a 2% disadvantage in the next, produces a net zero. Over 240 months, the residual signal is in the third or fourth decimal of return. By 360 months, it is barely visible at all.
Meanwhile, the signal — the long-term path of the index, compound returns of 10–13% per year — is multiplying the entire corpus by 7x to 18x over the same horizon. The signal is enormous; the day-of-month noise is rounding error against it. The compounding engine wins not by being clever, but by being huge.
This is the Coil Principle applied to the calendar. Long horizons accumulate so many units that the precise NAV of any single buy stops mattering — the population of units defines the corpus, not any individual one. The Coil Principle → develops this dynamic in detail.
If day-of-month doesn't matter, who cares whether someone obsesses over it? They do, and not just because the obsession itself is a small cognitive tax. The real cost is not the optimisation — it's the delay it produces.
This is the structural shape of the trap: the cost of optimising the non-variable is paid in the variable that does matter. Days deliberated turn into months delayed turn into compounding forfeited. The investor who agonises over which day to start ends up starting later. The one who picks any day at all and starts immediately ends up materially wealthier.
Once the investor has been freed from the day-of-month puzzle, attention can be redirected to the variables that actually drive long-term outcomes. These are the parameters that, when changed by the same proportion, change the final corpus by orders of magnitude more than day-of-month does.
The flagship interactive tool runs the 30-investor experiment on real Nifty 50 prices. Pair with companion calculators to see what the variables that do matter look like quantitatively.
"There is a question that feels like a decision and is not. It is the day of the month on which the SIP debits. First, fifth, tenth, twentieth. The forums fill with theories. The investor postpones. Twenty years of real Nifty closes give the same answer regardless of which day was chosen: the corpora differ by less than one month of one investor's SIP. The compounding engine is too large for the calendar to matter. The choice that feels like the answer is part of the question. The variable that moves the outcome is whether the investor started, stayed, and continued. Optimise that. Pick any day. Begin."