How we count government tenders

5 min read · Last reviewed 14 August 2026

Tenders Agent's tender counts are produced by reading each government procurement portal directly, consolidating the notices into a single corpus refreshed daily, and classifying every tender into a fixed 40-category sector taxonomy and a canonical state list — so that counts are comparable across portals that use incompatible category schemes of their own.

Any number about the tender market depends entirely on what was counted and how. This page states both, so that the figures in this section can be assessed rather than taken on trust.

Where the data comes from

We read each procurement portal directly, rather than buying an aggregated feed, and consolidate the notices into one corpus. The sources are:

• GeM (gem.gov.in) — the Government e-Marketplace: central ministries, PSUs and state bodies buying goods and services, as both bids and reverse auctions.
• eProcure / CPPP (eprocure.gov.in) — the Central Public Procurement Portal, covering central ministries, departments and public sector undertakings.
• CPPP ePublish — notices central organisations publish without running the bid online. They never appear on the eProcure tender list, which makes them the most commonly missed category of opportunity.
• Coal India (coalindiatenders.nic.in) — Coal India and its subsidiaries on one portal.
• Maharashtra (mahatenders.gov.in) — state departments, municipal corporations, zilla parishads and boards.
• IREPS (ireps.gov.in) — the Indian Railways e-Procurement System: works, goods and services tendered by zones, divisions and production units.

The corpus is refreshed daily. A tender enters it once its notice is published on a source portal, and stays in it after closing — the archive is retained rather than pruned, so historical tenders remain readable.

How tenders are classified

The portals do not share a category scheme, and several of them use free text where a controlled vocabulary would be expected. Counting across them therefore requires re-classification rather than aggregation.

1. Sector. Every tender is classified into one of a closed 40-category taxonomy — 26 goods categories, 13 service categories and a fallback bucket for notices that genuinely do not fit. The taxonomy is fixed rather than emergent, so counts are comparable over time and between portals.
2. State. State and union territory names are normalised to a canonical list of 36. The raw corpus carries legacy and variant spellings — Orissa for Odisha, Pondicherry for Puducherry, abbreviations, inconsistent use of 'and' versus an ampersand — which are mapped to the canonical name before counting.
3. Value and dates. Estimated value, EMD, publication and closing dates are taken from the notice and its documents where present, and left empty where the source does not state them. They are not imputed.

What 'open' means

An open tender is one whose closing date has not passed. Counts described as open exclude closed tenders; counts described as total or corpus-wide include the archive. Where a page states a figure, it states which of the two it is.

Known limitations

These are real and worth knowing before quoting a figure:

• Coverage is portal-bound. A tender published only on a portal we do not yet read is not in the corpus. We cover the national portals and a growing set of state and PSU systems, not every procuring authority in India.
• Classification is automated. Sector assignment is performed by a model against the fixed taxonomy, not by hand. It is accurate in aggregate and occasionally wrong on an individual notice, particularly where a title is uninformative.
• Some notices carry no value. Where a portal does not publish an estimated value, we leave it empty rather than estimating one — so value-weighted figures describe only the subset that stated a value, and are not a measure of total market size.
• A minority of notices lack attached documents. Where a source portal did not expose a downloadable document, the tender's structured detail is thinner than the rest of the corpus.
• Counts move daily. Tenders publish and close continuously, so a figure described as live is accurate as at the moment the page was generated and will differ tomorrow.

Using these figures

The counts are most useful comparatively: which sectors carry sustained volume, which states are most active in a sector, and where a supplier's capability meets a market that is actually tendering. They are least useful as an absolute measure of the size of Indian public procurement, which is a question about spending and belongs to the government's own reporting.

Key takeaways

• Figures come from reading six procurement portals directly, refreshed daily.
• Sectors use a fixed 40-category taxonomy; states are normalised to a canonical 36.
• Open means the closing date has not passed; the archive is retained separately.
• Coverage is portal-bound and classification is automated — both have known error modes.
• These are counts of tenders, not of awards or of expenditure.

Frequently asked questions

Where does Tenders Agent's tender data come from?

From reading government procurement portals directly rather than from an aggregated feed. The sources are GeM, the Central Public Procurement Portal (eProcure), CPPP ePublish, Coal India's portal, Maharashtra's state portal and Indian Railways' IREPS. The consolidated corpus is refreshed daily.

How are tenders classified into sectors?

Every tender is classified into one of a closed 40-category taxonomy — 26 goods categories, 13 service categories and a fallback bucket. Classification is automated against that fixed list, which makes counts comparable across portals that use incompatible category schemes of their own.

How many government tenders are open in India right now?

The live count across the portals we read is shown on the Tender Market Reports page and is generated at request time. It counts tenders whose closing date has not passed, across GeM, CPPP, CPPP ePublish, IREPS, Coal India and Maharashtra — not every procuring authority in India, and not contract awards or expenditure.

Do these figures measure government procurement spending?

No. They count tenders — opportunities open to bid on — not awards made or money spent. Many notices do not publish an estimated value at all, and those are left empty rather than estimated, so value-weighted figures describe only the subset that stated a value.