Who is actually buying, and what the print cannot see

What the number is

After the close, one line is published for each institutional category: total buy value minus total sell value in the cash segment, in ₹ crore, for that session. FII is the foreign institutional category, DII the domestic one. That is the whole number — a subtraction, on one segment, for one day.

Everything difficult about reading it follows from how little that definition contains.

The labels have some history. The foreign institutional investor route was folded into the wider foreign portfolio investor framework years ago, so the accurate modern term is FPI; market usage kept saying FII and the two now travel together. Our own feed treats them as one category, which is why the panel reads FII / FPI rather than picking a side.

The panel shows three cells — the foreign net, the domestic net, and the two added together — against the date of the session they belong to. That date is worth looking at. Flow is a settled, end-of-day figure published after the close, so a panel open during market hours is carrying the previous session. The cross-asset tiles beside it move while you watch them; this one cannot.

‘Net’ hides more than it shows

A net figure is a difference, and a difference throws away both of the numbers that produced it.

Take an illustrative session in which the foreign category buys ₹9,000 crore of shares and sells ₹9,200 crore. The print is −₹200 crore, and it reads as a quiet day. It was not a quiet day — nearly ₹18,000 crore of stock changed hands inside that one category — but the report was never built to say so. A small net is consistent with a violent session and with a dead one, and nothing on the panel distinguishes them.

The netting also runs across desks, which is the part that gets forgotten. "FII" is not an investor. It is a bucket containing an index fund tracking a benchmark, a long-only manager cutting an overweight, a hedge fund putting one on, and a sovereign fund doing something unrelated to any of them. When one of those buys what another sells, the bucket nets to roughly zero and the panel reports agreement where there was a transaction between two disagreeing parties.

So the honest reading of a small print is the category did not move much on balance, which is a weaker statement than it looks. Foreign investors were not calm. Foreign investors, added up, were flat.

The word doing the most work is ‘cash’

The figure covers the cash segment: shares bought and sold on the exchange, delivered and settled. That single qualifier decides most of what the number can support.

Indian institutions also hold exposure through index and stock futures, through options, and through primary-market participation. None of that is inside the cash print. A desk can be a net seller of shares and, in the same session, hold or add a long index futures position — and the flow panel will show only the selling. Nothing is being concealed; a cash-segment report cannot contain a derivatives position any more than a bank statement can contain a mortgage.

Here is what the panel can and cannot answer, question by question.

The specific mistake this table exists to prevent is the sentence "foreign investors pulled ₹X crore out of India today". The figure supports "the foreign category was a net seller of ₹X crore of shares in the cash segment". Those are different claims, and the gap between them is where most flow commentary lives.

Where domestic money comes from, and when it arrives

The foreign figure and the domestic figure look symmetrical on the panel. They are not the same kind of number, because the money behind them does not arrive the same way.

Sitting inside the domestic category are mutual funds, insurers, banks and pension money. Much of what those institutions deploy into equities is not a decision taken that morning. It is money that landed because a standing instruction fired: a monthly systematic instalment — a SIP — debited from a salary account, an insurance premium, a payroll retirement contribution. A fund manager who receives subscriptions into an equity scheme has a mandate to be invested; the money has to go somewhere, and the mandate, not the news, decides roughly where.

So a large part of domestic institutional buying is keyed to the calendar rather than to the market. The instruction that produced today's purchase was set up months ago by someone who has not looked at a screen since. That is a mechanical claim about how the money gets there, and it is the single most useful thing to know about the DII line.

It does not make the domestic figure inert. Redemptions run the other way, fund managers hold and release cash within their limits, and insurers and pension funds make genuine allocation calls. The scheduled part is a floor under the flow's behaviour, not the whole of it — and the part that is behavioural shows up when instalments get paused during a fall, which is its own subject in what stopping a SIP in a drawdown actually does.

The trade-off in reading it this way: recognising the scheduled component makes the domestic number less informative as a sentiment signal, and more informative as a structural one. You lose "domestic investors are bullish". You gain an accurate description of the character of that demand — price-insensitive, recurring, and indifferent to the headline that moved the foreign side.

The mirror is partly an accounting identity

Now the observation that produces more bad analysis than anything else on the page: on many sessions the two figures are near mirror images. Foreign selling of roughly the size of domestic buying, day after day.

Start with the arithmetic, because part of the answer is there before any behaviour is involved. Every share sold in the cash segment was bought by someone, so summed across every participant in the segment the nets cancel to zero. The published participant categories divide up that same market: foreign and domestic institutions are two of them, and retail, proprietary desks and corporates are the rest. It follows that whenever the rest of the market is roughly flat on balance, the two institutional nets must be close to opposite in sign and similar in size. Not because either side responded to the other. Because the numbers have to sum to zero and the other terms were small.

That is the identity. Layered on top of it is the schedule from the previous section: the domestic instalment flow arrives whether or not anything happened, so it is available to be on the other side of foreign selling without anyone having formed a view.

Put the two together and you get a named failure mode — call it the absorption story. It runs: foreign institutions sold, domestic institutions bought, therefore domestic money is defending the market. The panel does record that both things happened. What it cannot establish is that the second happened because of the first, and both the identity and the schedule give you reasons to think that much of the time it did not. Two unrelated processes met at a price, and the report of the meeting reads like a conversation.

None of this says the domestic bid is unreal or that the symmetry is meaningless. It says the symmetry is weak evidence of intent, and it is routinely quoted as strong evidence of intent. The same discipline applies to the correlation heatmap elsewhere on the page — a measured relationship is not a mechanism, a point the pillar on reading macro signals works through in full.

What our model does with the number, and the ceiling it puts on it

Our composite macro score folds foreign flow in with a weight of 0.12, saturating at ±₹5,000 crore of net foreign cash. Both figures are FNOTrader design choices — considered judgements about what matters to Indian equities, not measured constants and not the output of any estimation. A different reasonable view would pick different ones.

Three properties of that choice are worth stating, because each changes how the number should be read.

One: past the saturation point, a bigger number is not a bigger signal. The input is the net divided by ₹5,000 crore and then clamped to the range −1 to +1. Illustratively: net buying of ₹2,500 crore contributes half of the input's maximum, ₹5,000 crore contributes all of it, and ₹15,000 crore also contributes all of it. A record session and a merely large one enter the score identically. That is deliberate — it stops one extraordinary day from dominating a cross-asset average — and the cost is exactly what it sounds like: the score cannot tell an unusual flow day from an unprecedented one.

Two: only the foreign figure scores. The domestic net is fetched, shown on the panel and added into the displayed total, and it carries no weight in the composite at all. The reasoning is the schedule argument above: the score is trying to summarise the global risk environment, and a number substantially produced by standing instructions in Indian salary accounts is not a reading on that environment. This is our judgement and it is contestable — someone building the same score for a different purpose would weight the domestic line, and would not be wrong to.

Three, and this is the part worth doing the arithmetic on: the flow input cannot move the regime by itself. The tile table carries twenty weighted inputs summing to 1.33; flow is not one of them, and the code adds its 0.12 to the numerator and the denominator at scoring time. So when every feed reports there are twenty-one inputs and the divisor is 1.45. Flow's share of that divisor is 0.12 ÷ 1.45, or about 8.3% — which means a fully saturated flow reading, with every other input reading zero, produces a score of 8 on a scale of −100 to +100. The risk-on and risk-off labels sit at ±20. With the rest of the board reporting, a record foreign flow day does not reach them on its own.

Two smaller mechanics complete the picture. The divisor falls when a feed fails, because a missing input leaves the numerator and the denominator together — 1.45 is the all-present maximum, not a constant, and every surviving input's share rises when something drops out. That is the caveat on the 8.3% above: it is flow's share of a full board, and a half-dead board gives whatever survives more of the score. And the flow term is included whenever either figure is non-zero, so a session with domestic flow and an exactly zero foreign net still consumes its 0.12 of the divisor at a contribution of zero, nudging the score a little toward neutral.

Flow has no tile of its own, and the reason is a unit mismatch rather than a view about importance: every tile reports a percentage change measured against its own scale, and flow is an absolute rupee figure. So it gets its own three-cell panel instead — and it can still appear by name among the rationale lines under the gauge on a day when it lands among the four largest contributors.

Four ways the flow print gets misread

1. Reading a cash sale as an exit. The figure covers one segment. A net seller of shares can be adding exposure in index futures in the same session, and neither the addition nor its size appears anywhere on this panel.
2. Reading the mirror as a duel. The absorption story. Part of the symmetry is the sum-to-zero identity and part of it is the instalment schedule; what is left over as evidence of intent is much smaller than the chart suggests.
3. Reading a small net as a quiet day. Gross activity is not in the figure. Two large opposing desks inside the same category produce a print near zero.
4. Reading flow as a reason to act. A settled figure describing the previous session tells you about the environment a position sits in, not about the position — it has no view on your holding period, your allocation or the price you paid. Those belong to asset allocation, which is a different question with a different timescale.

There is a fifth that is less a misreading than a missing leg. Foreign flow is the visible end of a chain that starts with global funding conditions — the dollar, the rate an allocator earns at home, and the currency risk of coming here. The flow print is where that chain surfaces with a rupee sign attached, which is why it feels like the most concrete number on the page and why it is the last link rather than the first. The upstream legs are in the dollar index and emerging markets and US yields and Indian share prices, and the leg where a foreign investor's return is converted back out of rupees is in USD/INR and your portfolio.

Where this sits in the app

The flow panel described here is on the macro page in FNOTrader's Options Analytics app: the foreign net, the domestic net and their total in ₹ crore, dated to the session they belong to, alongside the cross-asset tiles and the composite score.

The weight of 0.12, the ±₹5,000 crore saturation, the exclusion of the domestic figure from the score and the ±20 regime cut-offs are the ones stated in this article. They are published rather than buried because a reader who disagrees with a weighting should be able to see precisely what they are disagreeing with, and recompute the number their own way.

Common questions

Continue reading

More in Macro & Markets · App: Options Analytics · Definitions: glossary · Free tools: calculators · All: every article