CAGR, XIRR and rolling returns: three answers to three different questions
Three numbers, three different questions. Most arguments about fund returns are really arguments about which of the three somebody quoted.
A fund has one history and at least three legitimate ways to describe it. They disagree, and each of them is right about a different question.
CAGR: what one lump sum did
Compound annual growth rate answers: if I had put money in on this date and left it, what steady annual rate would have taken it to today's value?
It needs exactly two facts — the value then and the value now — and it smooths everything in between into one rate. A fund that fell 40% and then tripled shows the same CAGR as one that crept up every year, if they arrive at the same place.
What it cannot tell you: anything about money added along the way. If you invested monthly, CAGR describes an investment you did not make.
XIRR: what your actual instalments did
XIRR answers: given every rupee I put in and the date I put it in, what annual rate makes all of that arrive at today's balance?
A SIP started three years ago has 36 different holding periods — the first instalment has had three years to work, last month's has had one month. XIRR weighs each by how long it has been invested.
This is why a SIP's XIRR and the fund's CAGR are rarely the same number, and why neither is lying. In a market that rose steadily the CAGR will usually look better, because your later instalments did not get the early run. In a market that fell and recovered, the SIP often looks better, because the later instalments bought cheaply.
If you invest monthly, XIRR is your number. CAGR is the fund's.
Rolling returns: taking the start date away
Both of the above depend enormously on one arbitrary thing — the date you started. Pick a start just before a crash and any fund looks poor. Pick one just after and it looks excellent. Neither says much about the fund.
Rolling returns remove that. Instead of one five-year period, they compute every five-year period in the fund's history — start in January, start in February, start in March — and show the spread.
What comes out is more honest and less comfortable: not "this fund returned 14%", but "across every five-year stretch it has lived through, this fund returned between 4% and 21%, and was above 10% in most of them".
That range is the thing worth knowing. It is the closest an ordinary number gets to telling you what the experience of holding something has been, rather than what one lucky window looked like.
Which to read, and when
- Comparing two funds? Rolling returns, over the same window. Single-period CAGR comparisons are mostly comparisons of start dates.
- Judging your own SIP? XIRR. Nothing else accounts for your instalments.
- One lump sum, one date? CAGR is exactly the right tool.
One honest warning about all three
Every one of these describes what already happened. None of them is a forecast, and a fund that produced a high number over one period has no obligation to produce it over the next. The arithmetic is solid; the extrapolation is not, and the extrapolation is the part people do in their heads.
On this site, the comparison tool shows CAGR, XIRR and rolling returns for the same fund over the same window, so you can see how far apart three correct answers can be.
Analysis, not advice. Past returns, however they are measured, do not tell you what comes next.