The factory, the shop, and the line between them.

The ASI is one survey published under four industrial classifications, and each classification is a different ruler. The lines are deliberately never joined: at every seam the survey re-sorted its factories, and a step across a seam measures the re-sorting as much as the economy. Read each segment's slope, not the jumps between segments.

What the segments say is uncomfortable enough. The register drifted through the 1990s, fell in the early 2000s, and the people-panel is worse: factory employment went backwards for half a decade after 1997-98: the famous lost years of organised manufacturing, when output rose and jobs did not. The climb since is real; the 2010-11 step inside the current block is the frame revision, annotated in Plate 2.

Counts are the honest part of the ASI: a factory is a factory in any year's rupees. The one violent move in the upper panel is 2010-11, when the survey's sampling frame was rebuilt and tens of thousands of already-existing factories walked into view at once. The lower panel barely flinches at the same moment: the newly visible factories were small, so they carried few people with them.

Everything else is slow. The register grows faster than the count in operation; the gap between the two lines is factories that exist on paper but reported no activity.

Every rupee here is a rupee of its own year. The upper lines mix real expansion with fifteen years of inflation, and the survey publishes no deflator to separate them: so this plate states levels and refuses to call them growth. The dashed panel below is the honest companion: year-on-year change, which at least shows the rhythm: a pandemic dip, then a violent rebound that is part recovery and part the price surge of 2021-22.

The steadier story is the gap between the lines: what factories add and what they pay out as wages and salaries. Emoluments have stayed near a third of value added through the whole window.

Factory India is coastal and western. The long tail is the plate's second story: below the top ten states, shares fall under two per cent each, and the whole north-east together rounds to almost nothing. Concentration this steep is old, ports, power and supplier networks compound, and it moves only slowly, survey by survey.

Share of value is not share of factories. Tamil Nadu runs the most factories; Maharashtra makes the most value per state. Capital-heavy industry, not counting, decides this ranking.

Each panel is one industry's sixteen years of nominal value added, drawn to its own scale: the grid compares trajectories, not sizes; the printed share does the sizing. The top row is heavy industry and process plants: metals, vehicles, chemicals, food, pharmaceuticals. The famous labour-intensive exporters, textiles, apparel, leather, sit far down the shelf, large in employment and modest in value added.

Nominal caution applies panel by panel: a rising line is prices as much as product. The shapes still discriminate: some industries dip in 2020-21 and snap back, others never quite do.

One dot per industry, plotted twice: once for its share of all factory jobs, once for its share of all factory value added. If jobs and value tracked each other, every dot would sit on the diagonal. They do not. The dots below it carry more workers than their output warrants: the labour-intensive backbone. The dots above carry more output than their headcount: the capital-intensive core.

The value-added grid above ranks industries by money; this strip ranks the same industries by people. The two rankings do not agree. Textiles, apparel and food, the classic labour-intensive trades, climb to the top here while sitting far down the value shelf. Basic metals and chemicals, the value leaders, carry fewer people than their rupee share implies.

Read this plate against the scatter above it: the scatter plots both shares at once and names the gap; this strip puts employment share on its own axis, so the ranking is legible end to end.

Divide the year's wage bill by the number of people on the rolls and you get what an average job in that industry is worth. Divide by twelve for a month. The low end is the work India does by hand: rolling bidis, cutting and stitching cloth, tanning leather, sawing wood. The high end is capital-heavy plants where a few people run expensive machines, and where a large share of the bill goes to engineers and managers. The same factory floor can hold both: these are averages over everyone, from the helper to the plant head.

The Factories Act starts counting at ten workers with power, twenty without, so even the smallest unit here is a real workplace. Car plants, electronics assembly and pharmaceuticals run large shifts under one roof. Rice mills, sawmills and cotton gins are small and scattered. Put this plate next to the one above and size and pay do not line up: bidi and garment factories are large and pay the least, while refineries are middling in size and pay the most. What a job pays follows the machines behind each worker more than the headcount around them.

The two-digit divisions above lump together very different trades: "food products" holds rice mills, dairies, biscuit factories and edible-oil refineries. The four-digit class is the finest cut the survey publishes, close to how a factory owner would describe the business. A few classes, refining, steel, pharmaceuticals, cars, carry a large share of the value, and they are mostly capital-heavy: compare each bar with its share of jobs.

Read across a row to see what a state's factories make; read down a column to see where an industry lives. Steel needs iron ore and coal, so it sits where the mines are; car plants gather in clusters of suppliers around a few cities; medicines and chemicals follow ports, skilled labour and decades of policy. A state that leans on one industry rises and falls with it: a bad year for steel prices weighs on Odisha's factory accounts more than on any other state's.

This is the guide's one picture worth arguing about. A year of work in a registered factory yields several times the value added of a year in an unincorporated establishment. The gap is real, and it is not a verdict on the people: the factory worker stands next to machines, capital and scale that the shop worker does not have, in industries the shop cannot enter. The bars measure the establishment, not the person in it.

Read the two pale bars against each other too: even inside the informal economy, an establishment that hires at all adds roughly double per head what an own-account one does. Hiring one person is India's steepest step up.

Each row is one activity; the line connects the rural value (grey dot) to the urban value (blue dot). The top panel counts workers; the bottom panel values them. Read the two panels together: where the workers sit and where the value sits are different answers.

Manufacturing workers skew rural: the village workshop, the handloom. Trade and services skew urban. But GVA per worker skews urban in every activity: the city establishment adds more per head than the rural one, whether it makes, trades or serves. The gap is widest in manufacturing and narrowest in services.

The factory GVA strip earlier in this guide concentrates in three western states. This mirror, informal workers, not factory value, spreads wider. Uttar Pradesh, West Bengal and Maharashtra hold the largest shares, but the tail is long: states that barely register in factory output carry millions of informal workers each.

The comparison is the point. A state's rank here against its rank on the factory strip measures how much of its economy the ASI misses: and for most states, that is most of the economy.

These are the shops, stalls, workshops and service counters that are not companies: the kirana store, the tailor, the auto-rickshaw owner, the tuition centre, the roadside dhaba. Each bar counts the people who work in them, owners and helpers included. Retail is the giant because a shop is the easiest business in India to open: little capital, often no registration at all, frequently run from the front room of a house.

Much informal work done by women happens at home: rolling bidis, stitching garments on piece rates, weaving, running a tuition class from the verandah. Work outside the home, on the road or in the repair shop, is almost entirely men's. The survey counts a woman as a worker in an enterprise only if she works in it; women who work on the family farm or only at home for the household are outside this count altogether.

The informal economy is where India's less-favoured groups own things. The owner distribution leans further toward OBC and SC proprietors than the ownership of registered companies ever has, and that is the quiet function of this sector: it is the open door. The bar below it shows the other gate: sex. Most establishments are one proprietor, and most proprietors are men; female-owned establishments cluster in manufacturing, largely home-based work.

Education completes the portrait: the modal owner has schooling between primary and secondary. Graduates are a minority of owners here: the graduate premium in India is a wage premium, collected in jobs, not shops.

Two Indias share this chart, and one surprise. Bank accounts are the reached frontier: three establishments in four keep one, a decade of financial-inclusion policy visible in a single bar. The computer never arrived: a few dozen per thousand, flat across all three rounds. The surprise is the internet bar between them, which nearly doubled in two years: the smartphone became a business tool without a computer ever entering the shop.

The registration bars are the formality frontier. An establishment registered under the Shops Act or GST is still unincorporated: registration is a spectrum, not a switch, and most of this economy sits at the unregistered end of it.

The picture of precarity that "informal" suggests, a street vendor, seasonal, a few hours a day, is the minority case. Most establishments have a roof and an address: household premises or a permanent structure outside one. Most run more than nine months a year, and the modal working day is a full eight-to-eleven hours. Informal describes the paperwork, not the work.

The margins are still where the hardship lives. The mobile and street-vending sliver at the end of the top bar has no premises to return to, and the short-months establishments at the left of the middle bar are the seasonal trades: brick work, sugarcane crushing, festival goods.

This is the survey's whole balance sheet, in the survey's own order: capital first, then output and inputs, the value-added lines, the people, the wage bill, and fuel at the end. The plates above chose eleven of these lines; the other forty-six are here so that the choosing is inspectable. A reader who wants depreciation, mandays, contract workers or the coal bill does not need another source: it is all one table.

Everything monetary is nominal, and the sparklines mostly rise for that reason alone. The exceptions earn attention: a line that falls in nominal terms, rent paid for fixed assets, unpaid family members working in factories, fell in fact. One line is missing by design: the children-workers indicator is served only as a single corrupt row and is quarantined, not printed; 56 of the 57 lines are here.