GeM - India Government e-Marketplace

India buys its government equipment through one front door. Government e-Marketplace (GeM) is a wholly government-owned Section 8 company under the Department of Commerce, and it is where ministries, departments, PSUs and other public buyers procure common-use goods and services. The scale stopped being arguable years ago: 137,779 primary buyers and 214,559 secondary buyers operate on the platform across 10,649 product categories and 350 service categories, together placing 7.58 million orders worth Rs 5,027.78 billion in the last completed fiscal year - with 2.30M orders already logged in the fiscal year underway at the August 2026 research pass.

For heavy electrical equipment the relevant fact is that this is where public demand actually clears. The portal's own category navigation carries an Electrical Appliances cluster - molded-case circuit breakers, LED luminaires, cabling - and heavy items such as power and distribution transformers are procured through GeM bids and direct purchase rather than through paper-era tender channels.

Three surfaces make up the record. Individual contract orders carry the administrative, buyer-hierarchy and line-item fields described below. Bid listings expose the contested side of buying, searchable across product, service, BOQ, rate-contract and limited-tender types with high-value and evaluation-status flags. And a procurement-statistics layer publishes per-financial-year counts and values for tenders published, price bids opened and contracts awarded - split by Central Government, CPSUs and States, and by Goods, Services and Works - with FY2024-25 showing 1,199,910 central tenders published. Datadory packages all three as typed, joinable records under Heavy Electrical Equipment.

One row per contract order, flat enough to load without a parser. Values below are exactly as observed during the August 2026 verification pass:

# one row per contract order - captured during the August 2026 verification pass

contract_no       : GEMC-511687701075757
order_status      : Fullfillment in Progress        # portal spelling, shipped verbatim
organization_type : Central Government
ministry          : Ministry of Agriculture and Farmers Welfare
contract_date     : 03/3/2018 12:18
buying_mode       : Direct
total_value_inr   : 10692.000
product           : Raisco Basketball Ring (5 Basketball Size With Net)
brand             : Unbranded
hsn_code          : -
ordered_quantity  : 1
unit_price_inr    : 10692.000

contract_no       : GEMC-511687768833836
order_status      : Order Rejected by Seller
organization_type : Central Government
ministry          : Ministry of Agriculture and Farmers Welfare
contract_date     : 09/2/2018 10:12
buying_mode       : Direct
total_value_inr   : 27040.000
product           : Dell Vostro 3268 SFF - PDC G4560-4GB-1TB-NO DVD-18.5" -Linux - 3Y - WIFI - BT
brand             : Dell
model             : Raisco Basketball Ring
hsn_code          : 84714900
ordered_quantity  : 1
unit_price_inr    : 27040.000

Two honest observations about what you just read. First, the portal spells "Fulfillment" as "Fullfillment" - delivered values keep the original spelling, because silently correcting status strings is how joins quietly break. Second, the observed rows sit outside the electrical cluster, and that is the point: the row shape is identical in every category. The electrical slice - transformers, switchgear and circuit breakers, cables, luminaires - rides the same fifteen columns, routed by product text and hsn_code rather than by a separate schema. Your sample pulls records scoped to the families and buyer groups you name, with the column set held steady across every delivery.

Fifteen fields hold the record, and they divide into three jobs. An administrative layer (contract_no, contract_date, buying_mode, order_status) makes each order addressable and records how it was won - direct purchase versus bid-backed award - and whether it actually fulfilled. A buyer-hierarchy layer (organization_type, ministry, department, organization_name) resolves who is spending, from the Central-Government-versus-State split down to the specific office that clicked the order button. A commercial layer (product, brand, model, hsn_code, ordered_quantity, unit_price_inr, total_value_inr) carries what was bought, how many, and at what price per unit - which is the layer every pricing and market-sizing question lives in.

Every definition in the dictionary below was checked against the rendered contract-record presentation during the August 2026 research pass rather than inferred from column headers, which is why the examples carry real observed values. Identifiers ship verbatim, so delivered rows reconcile cleanly against anything already collected on your side.

• Geography: the whole of Indian public buying. All central government ministries, CPSUs and states transact on the platform, and the buyer-hierarchy fields preserve which tier placed each order - so central-versus-state demand comparisons need no extra geocoding work.
• Temporal: the portal has been live since 2016, and contract records were observed running from February-March 2018 onward during the research pass. The aggregate statistics layer is organized around the Indian financial year (April to March), with recent years selectable, so fiscal-calendar analysis is native rather than retrofitted.
• Granularity: two grains, deliberately. The order level gives you individual contracts with line items underneath - the grain at which pricing questions get answered. The financial-year level gives you published counts and values for tenders, price bids and awards, split by buyer class and by Goods, Services and Works - the grain at which headline demand arguments get made. Holding both lets an annual claim be audited downward to the rows that produced it.

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

An order stream measured in millions of rows per fiscal year rewards scheduling. Take the historical contract base once - every contract number, buyer tier, buying mode, line item, quantity and price as typed columns - then keep new orders rotating onto whatever cadence your systems expect. Weekly suits market-sizing and account-planning work, where the quarter matters more than the day; daily fits competitive workflows, where a large transformer or switchgear award is news the moment it lands. The cadence setting governs how fast an order reaches you, never what its columns mean. Either way, a sample goes out first - cut to the equipment families and buyer groups you name. Never a screenshot; always rows.

Six jobs the record settles outright:

1. Supplier win and price analysis - who wins transformer, switchgear and cable orders, at what unit price, through which buying mode; see competitive intelligence workflows built on awarded lines.
2. Public-demand market sizing - bottom-up totals for Indian government demand for heavy electrical equipment, defended against the financial-year aggregate tables; see market sizing.
3. Account prioritization - ministries, departments and offices ranked by electrical purchase volume and growth before sales resources commit.
4. Price benchmarking - quoted tender prices tested against awarded unit prices for equivalent goods, scaled by ordered quantity within the same buyer tier.
5. Procurement-policy research - direct-versus-bid mix, status distributions and rejected-order share read as evidence under India's General Financial Rules discipline; pairs naturally with citation-grade research standards.
6. Demand forecasting for capacity planning - order flow by financial year and buyer segment feeding production planning for manufacturers selling into the public sector.

Each job runs off the same fifteen columns; the difference is which cut you take and how deep the line-item expansion goes.

Competitive intelligence and product teams get the win-loss picture reconstructed from awarded lines - who bought what, at which unit price, through which buying mode; see competitive intel product teams use cases. Market researchers and consultants get a bottom-up size of public-sector demand, citable down to individual contract numbers; see market researchers use cases. Data scientists and ML engineers get millions of order rows on stable GEMC keys with a categorical spine ready for pricing and forecasting models; see data scientists use cases. Journalists, academics and students get procurement claims that land on a named contract, ministry and date rather than an unnamed statistic. Developers and data-product builders get schemas that hold steady between pulls; see developers builders use cases.

This record owns the transaction: the moment public money changes hands for equipment, with buyer, terms and line-item prices attached. Its neighbors own different territories. The Ministry of Heavy Industries - India Heavy Electrical Vertical covers the industry and policy frame - capacity, schemes, sector statistics - without telling you who bought a transformer last week. Marketplace catalogs cover supply, not public demand: IndiaMART Transformer & Electrical Suppliers and Alibaba Transformer & Electrical Equipment Listings show what sellers offer and ask, which prices only become real when an order awards. The data.gov US Federal Transformer Datasets record is the American public-sector counterpart, and T&D Europe Commodity Indices & Reports adds the input-cost pressure behind every quoted price.

Datadory rates this record 6/10 against a catalog mean of 7.81 across 1,744 datasets. The discount reflects the portal's own presentation quirks - rounded aggregate figures, free-text status values - not thinness of coverage; the order-level grain is the deepest public-demand view of Indian electrical equipment buying that exists.

Three things worth settling upfront.

First, scope defines the slice. Ten thousand-plus product categories means the undifferentiated firehose is mostly not your market. Naming the equipment families - transformers alone, switchgear plus cables, or the full electrical appliances cluster - and the buyer groups you care about turns the sample into a usable cut, and the HSN-code and product-text columns make the filter deterministic rather than keyword-fuzzy.

Second, counts move with the fiscal calendar. The 7.58 million order figure belongs to the last completed fiscal year, and 2.30M orders were already logged in the year underway at the August 2026 research pass. Any total you quote should state its census window, and delivered extracts report theirs rather than inheriting a number nobody maintains.

Third, decide the depth before the pull. Order-header rows answer who-bought-what-cheaply; full line-item expansion answers pricing questions properly at a larger row count. Both ride the same schema, so switching depth later costs nothing but the re-pull - tell us the question and the sample arrives shaped to it.