India Ministry of Statistics and Programme Implementation (MoSPI) Data

India counts itself here. The Ministry of Statistics and Programme Implementation, operating through the National Statistical Office, is the office legally charged with producing the country's official numbers - and its published estate is enormous: roughly 4,566 tables across 30 products in the e-Sankhyiki Data Catalogue, 187 unit-level microdata studies, and a machine-readable advance release calendar that schedules it all.

What sits inside matters far beyond the statistics community. CPI and CFPI print monthly for All India and every state and UT across Rural, Urban and Combined sectors - July 2026 put headline Rural inflation at 4.84 percent against Urban's 3.96 percent, a two-state economy hiding under one national number. National Accounts account for about 1,726 catalogue tables alone. Then the survey layer: ASUSE counting unincorporated establishments (roughly 2.48 crore in manufacturing in early 2026), PLFS carrying employment to 2025-26, HCES measuring what households actually spend, and the Annual Survey of Industries at factory grain.

One structural fact is worth knowing before building on it: the CPI basket was rebuilt at base year 2024, moving from 299 weighted items to 358 items across the COICOP-2018 hierarchy - 12 divisions, 43 groups, 92 classes, 162 subclasses. Every row carries base_year, so both regimes coexist in one table instead of silently overwriting each other.

Within Datadory's broadcasting data hub of 15 pooled datasets, this is the population-and-purchasing-power context any India-facing media-market model needs. Get a sample of this dataset cut to the states and series you actually model.

Three layers, three row shapes - price cells, survey observations, catalogue entries:

# layer 1 - CPI cell, monthly price index
base_year : 2024
series    : Current
period    : 2026 July
state     : All India
sector    : Rural
division  : CPI (General)
index     : 108.34
inflation : 4.84       # y/y, percent

# layer 2 - ASUSE observation, quarterly establishments
indicator  : No. of Establishments (in '00)
frequency  : Quarterly
period     : 2026 JAN-MAR
state/UT   : All India
activity   : Manufacturing
value      : 248493

# layer 3 - HCES consumption row
period         : 2023-24
state          : All India
indicator      : Average monthly per capita consumption expenditure (MPCE)
sector         : Rural
imputation_type: Without Imputation
value          : 4122
unit           : INR

Read them as one argument. The CPI rows say rural prices rose faster than urban ones in July 2026 - which means rural purchasing power erodes differently, which is precisely the input an advertiser pricing regional reach or a retailer sizing a store network needs before spending a rupee. The ASUSE row sizes the unincorporated manufacturing base; the HCES row says the average rural resident spent about 4,122 rupees a month in 2023-24. Demand, supply and prices, three rows.

The third dimension hides in plain sight: imputation_type. Survey rows distinguish imputed from un-imputed values, so a model can choose its own handling instead of inheriting somebody else's.

Sixteen fields cover the whole feed, splitting into three families that mirror the three layers above. The price-index family - base_year through inflation - identifies every CPI cell by base, period, geography, sector and COICOP division, then attaches the index level and the year-on-year rate. The indicator family - indicator_code, value, unit - carries the survey products, where the same three columns hold PLFS participation rates, ASUSE establishment counts and HCES monthly per-capita expenditure, discriminated by the unit column.

The catalogue family - table_name, frequency, release_date, unique_id - describes the publication layer itself. unique_id deserves attention: it combines product code, month-year and serial into a stable key like CPIMCY26001JULY, which makes 'has this table been revised since I last loaded it' answerable by comparison rather than by title matching.

A note on typing discipline: index values and inflation rates arrive as strings upstream and are cast to numeric on delivery against the declared types, so your warehouse receives typed columns rather than quoted numerics. Definitions below were verified field-by-field during the August 2026 research pass.

The sixteen-field spine covers every delivered row; four extensions fold into additional fields on request because each depends on which slice you name:

1. Survey indicator dictionaries. PLFS, ASUSE and HCES maintain separate indicator lists; the code-to-label crosswalk resolves against your sample.
2. National Accounts table schemas. Column layouts differ across the roughly 1,726 NAS tables by series and vintage; name the accounts and the schema is pinned.
3. ASI factory-record fields. Unit-level identifiers, NIC classification and state attributes ship as their own dictionary when factory grain enters scope.
4. Microdata study metadata. Study-level descriptors for the 187 unit-level datasets ride along where licensing scope is confirmed first.

None of these disturb the core keys, so widening the dictionary never breaks anything already built against it. Name your slices when you request the sample and the columns you tested are the columns you get.

• Geography: All India plus every individual state and UT. CPI covers all 36 states and union territories, each split Rural, Urban and Combined; ASUSE, PLFS, HCES and ASI are all state-breakable. A demand model for Bihar and one for Maharashtra draw from the same table.
• Temporal: CPI monthly from January 2011 on the 2010=100 base, with the 2012=100 back series running from January 2013 and the current 2024=100 series from January 2025. PLFS spans 2017-18 to 2025-26, ASUSE runs annually from 2021-22 and quarterly since 2025, HCES covers 2022-23 and 2023-24, ASI follows annual rounds.
• Granularity: item/sub-class/class/group/division depth for CPI by state and sector; household and person level behind the NSS-based surveys; factory level for ASI; establishment level for ASUSE.

The release rhythm is scheduled, not improvised: CPI prints at a fixed hour on the twelfth of each month or the next working day, governed by an advance calendar that itself arrives machine-readable. Delivery cadence from Datadory is decoupled from that schedule entirely - your feed moves when you want it to, not only when a press release lands.

API, files, or your warehouse. Daily, weekly, or hourly.

Scope decides the shape more than volume does. A pricing team usually takes the full CPI history once, then a monthly delta after each print; a market-sizing team takes specific NAS tables joined to state geography; a quant shop takes everything on a standing cadence beside its own macro panel. Whichever way you take it, the field dictionary travels unchanged, string-typed values arrive pre-cast to their declared types, and the sample precedes the contract - real rows first, commitment after.

• Media planners and advertisers convert CPI-by-sector into real purchasing power: a rural inflation print of 4.84 percent against urban 3.96 changes what a rupee buys where, and GRPs bought against nominal budgets misallocate accordingly. The market researchers broadcasting briefing shows where this slots into audience work.
• Investors and quant researchers treat the monthly prints as the coincident indicator of Indian household demand - consumption-expenditure levels from HCES calibrated against live CPI, national accounts for the long arc.
• Competitive-intel teams read ASUSE establishment counts by activity category as a private-enterprise heat map: where small manufacturing is forming, which states thicken fastest. The competitive intel broadcasting page has the workflow.
• Data scientists join the survey layer to their own transaction or viewership panels, using state x sector x period as the spine and the imputation flags to control data quality. See data scientists broadcasting.

What nobody should use it for: station-level or programme-level media facts. Those live elsewhere on the shelf - this record is the macro layer they all stand on.

Market researchers lead: no other record on the shelf ties Indian purchasing power to geography at this depth. Investors and quants come next for the same reason quants anywhere keep the national-statistics feed close - it is the denominator every private estimate gets judged against. Data scientists and engineers get clean join keys (unique_id, indicator_code) and pre-cast numerics, so the integration is a load job rather than a parsing project. Journalists and academics land last but not least: every number carries its release date and base-year regime, which makes citation precise.

If your work touches Indian consumers at all, the honest next step is a scoped sample - get a sample of this dataset and check the rows against figures you already trust.

Three things, stated plainly.

First, the base-year seam is real: 2010=100, 2012=100 and 2024=100 coexist, and the 2024 rebuild changed both the item list (299 to 358) and the classification tree (groups to the full COICOP division-class-subclass hierarchy). Splicing across the seam without respecting base_year manufactures false trends - we deliver the seam marked, not smoothed over silently.

Second, provisional prints revise. The latest month carries a provisional label in its own table title, and the catalogue unique_id lets you track a table across revisions deterministically. Build diffing on that key rather than on titles.

Third, scale varies enormously by layer: the CPI indicator side alone exposes well over a million rows while HCES is two survey rounds deep. Tell us which questions you are answering and the sample lands sized to those - not to the biggest table in the catalogue.

Notes worth reading before the rail:

• Provenance note - the underlying publisher is India's Ministry of Statistics and Programme Implementation through the National Statistical Office, the country's official statistical authority. Source terms are confirmed with you at sample request rather than asserted here.
• Freshness note - the catalogue held roughly 4,566 tables across 30 products at the August 2026 research pass, with the July 2026 CPI print already present. Delivery cadence stays decoupled from source behavior: hourly pickup simply re-delivers newest state between scheduled prints.
• Where to go next - the rail below collects the industry hub tying the shelf together, the neighbors covering audiences and stations at scale, and the head-to-head that settles when European media intelligence beats Indian national statistics and when you want both.

For vocabulary: what a consumer price index measures, and how microdata differs from the aggregated tables on this page.