What does a monthly portfolio not tell you?

An Indian mutual fund publishes the whole of its portfolio, line by line, every month. By world standards that is generous — plenty of markets make do with quarterly, and with the top ten only. It is also the hard limit on everything below, and on everything the rest of this site says.

How old is the newest portfolio anyone can see?

This is the first thing to do with any fund page anywhere, ours included: find the date on the portfolio before reading a word of it. A holdings table with no date on it is a holdings table you cannot use.

What falls between two photographs

HDFC Consumption Fund turned over 83% of its portfolio in a year, computed from 5 consecutive monthly disclosures. That figure is honest about what it measures and it is certainly too low, because it can only see the difference between one month-end and the next.

A stock bought on the 3rd and sold on the 20th never appears in any disclosure. Neither does a position doubled and halved back inside four weeks. What our turnover measures is net drift between disclosures, not trades — and that sentence belongs beside every turnover figure you will ever read, including the one the fund house prints itself.

It cuts the other way too. A fund forced to sell by redemptions is charged with churn it did not choose, and a fund whose weights moved because prices moved can look like it traded when it did not. A position's weight rises about as much as its price does, with nobody lifting a finger.

When the NAV and the book disagree

There is one way to see past the month-end, and it is arithmetic rather than information. Take the last portfolio a fund disclosed, carry it forward at each day's actual closing prices, and compare that to the NAV the fund published. The two should track. When they separate, something has happened that has not been disclosed yet.

The replication has its own error, and we say so rather than hiding it. A book with few priced lines, a stale close, a corporate action we have not adjusted for: any of these opens a gap that belongs to our arithmetic and not to the fund. That is why a fund whose replication we do not trust is marked noisy and is not used to teach anything, here or on its own page.

And some lines cannot be priced at all

Funds in this build have disclosed 1,869 companies between them. The exchange archive holds 5,807 equities in total, of which 2,609 trade only on the BSE — so most of the listed market is something no scheme here has ever reported owning.

A company that has delisted stops having a price. When that happens we show the date it last traded and print no price, rather than carrying the last close forward and quietly pretending the position is still worth something. A holding in a name with no usable price series cannot be marked to market, cannot enter a tracking calculation, and cannot be scored — so it is left out and said to be left out.

That is the rule the whole site runs on, and it is the only defence against the failure mode this lesson is really about: a number that looks complete because the missing part was silently filled in.

What this means for every measure here

• Turnover is a floor, not a total — it sees month-ends only.
• Holding persistence counts consecutive appearances in a top ten, so a position sold and bought back within a month reads as never having left.
• The top-ten hit rate judges positions from the date they were disclosed, which is after the manager bought them — part of the move is already gone.
• Every weight is a share of a book photographed at one instant, in a market that moved the next morning.

None of that makes the measures useless. It makes them measurements with a stated resolution, which is a different and more useful thing than a number with no error bars at all.