The monthly pulse of making things.

The long grey line is the retired base: a decade of single-digit hum, then the cliff of April 2020 (−57%) and the mirror-image spike of April 2021 (+134%): the two wildest prints in Indian statistics, both arithmetic as much as event. The blue line is the current base, begun April 2023. Where they overlap, they disagree by a point or two. That disagreement is not an error; it is what a new basket and a new panel measure.

Beneath the modern pair lie the three rulers India retired earlier: 1980–81, 1993–94 and 2004–05. Together they carry the monthly record back more than four decades: the steady eighties, the stall of 1991 when a balance-of-payments crisis choked imports of parts and fuel, the late-nineties grind, the 2003–2008 boom and the thin winter that followed. Where the lines overlap, from 1995 to 1998 and again from 2006 to 2011, they measure the same factories and disagree, sometimes by whole points. Neither is wrong; each holds a different basket.

Read these lines by the year, not the month. The item panels behind the old bases were smaller, so a single lumpy order could move an aggregate: capital goods once printed a 65 per cent year-on-year jump on one month's reading. The monthly sawtooth is arithmetic on a small panel; the slow drift underneath it is the economy.

A month is noisy; a year is not. The financial-year average smooths away festivals, strikes and one-off orders, and leaves the thing a household feels over time: whether there is more steel, cement, cloth and power in the country this year than last. Each point on a line is one year on one ruler. Where two rulers cover the same year they usually agree on direction and differ on size, because each weighs a different basket of goods.

A year of growth near 3 per cent and a year near 9 per cent are very different worlds for a factory worker: one is a plant running a single shift, the other a plant hiring for a second.

Each panel is one manufacturing industry on the retired 2011–12 base: the longest industrial memory the monthly statistics keep. The panels share one clipped scale, so a flat grey hum means a steady industry and a wide sawtooth means a lumpy one: a single large plant's shutdown can swing a two-digit industry by double digits. The lockdown spike is clipped out of every panel on purpose; it is the same arithmetic everywhere and teaches nothing about any one industry.

This grid is memory, not news. The current 2022–23 base lives in the plates above, and the two are never spliced: an industry's panel here ends where the old ledger ended.

Manufacturing is the heavyweight, most of the index, so the blue general line shadows the dark one. Mining and electricity swing wider and mean less individually. When the three point the same way, the print is broad-based; when they scatter, the headline is an average of disagreement.

Read the bars with their weights. Refinery products weigh 11.29 in the IIP, cement 2.16: a one-point slip in refining moves the index five times as much as the same slip in cement. Steel and cement rise together when India builds roads, railways and houses; crude oil and natural gas have fallen for most of the last decade because the old fields are ageing, which is why the country imports most of its oil.

The core index and the IIP move together because the core is two-fifths of the IIP, but the core swings less: power and fuel are bought every day, in good years and bad, while cars and machines can wait. When the IIP falls harder than the core, households and firms are postponing purchases, not switching off the lights.

Same index, different sorting: by what the output is for. Capital goods track the investment cycle: factories buying machines. Consumer durables track the confident household; non-durables, the everyday one. When capital runs ahead of consumer for a year, the expansion is investment-led, and the question becomes when the consumer joins.

The aggregate hides a split economy. In the same twelve months one industry can grow by a fifth while another shrinks, and the headline describes neither. The four lines are picked by the data each month: the two fastest and the two slowest over the last year. A single month moves them around; the twelve-month average is what earns a line its place.

The retired base's 23-industry ledger (§05) is a memory; this is the live field. Each panel is one industry on its own clock against a shared scale, ordered by the latest print. Read the rows, not the lines: where the current expansion sits, and where it does not reach.

The IIP counts goods. Services, more than half of everything India produces, had no monthly output measure at all: between one GDP quarter and the next, the largest part of the economy was statistically dark. The Index of Services Production is the answer, built like the IIP but counting bank transactions, freight, retail sales and the rest.

Read positions, not slopes. Each line is a level against its 2024–25 average: above 100 is above the base year, and that is all. The sawtooth is unmodelled seasonality: this index is too young for anyone to know what a normal October or a normal March looks like, so a month-to-month jump is not information yet. We do not know its seasonal shape. Nobody does; the history required to learn it does not exist.