Closed GeM tenders · 04 Dec 2023 – 08 Sep 2026

Credentials bind harder
than scale.

A supplier with a crore of turnover and the right authorisations can compete for more of the market than one with five crore and none. We read every attached specification and pre-qualification document — 1,50,086 of them — and measured which parameters actually determine who can bid.

This is a study of closed tenders, not a live feed. Every tender analysed here has already closed — the bidding window ran from 04 Dec 2023 to 08 Sep 2026, and 0 remain open. A further 26,497 tenders sit in the archive but have not closed yet, and are held out of every figure on this page. Nothing on this page is an alert or an opportunity. It is evidence about how tenders get written, used to coach bidders on what to look for next time. For help with a tender that is open right now, get in touch — that is a conversation, not a page.

What actually determines who can bid

The share of the corpus a supplier could have competed for, by profile. Turnover moves this far less than authorisation does: a supplier at one crore holding OEM letters reaches more of the market than one at five crore without them. Thresholds are the parameter most discussed; authorisation is the parameter that binds.

Value at stake. Of the contract value stated in these documents, 22,315 crore rupees (43%) sits in tenders requiring an OEM authorisation, and is therefore addressable only to authorised dealers. Bidders committed 1,683 crore rupees of earnest money against the tenders in this corpus. Neither figure implies anything improper — authorisation requirements are often necessary. They size the question of who the supplier base can include.

Access parameters, by prevalence

Share of tenders in which each parameter appears. Manufacturer-specific specification — the parameter most written about — is the least common. An OEM authorisation requirement appears roughly fourteen times as often and has a far larger effect on the qualifying supplier pool. Select any bar to see the tenders.

Ten observations, each one clickable

Every card opens the tenders behind the figure — buyer, deadline, thresholds, and which parameters were recorded. Nothing is modelled or inferred from titles; all of it is read from the document text.

Buyers apply different parameter mixes

Buyers differ less in how much they narrow a field than in which parameter they use to do it. Energy Department Uttar Pradesh operates largely through Proprietary Article Certificates; Urban Development UP through OEM authorisation. A single composite score would conceal that. Departments with 300+ tenders.

DepartmentTenders OEM authorisationProprietary articleNon-catalogue

Supplier Access Benchmark

A composite index across five parameter families that together determine how wide a supplier pool a tender reaches. It is peer-adjusted: each buyer is measured against the rate expected for the categories it actually purchases, not against the platform average — a fertiliser plant buying pumps is not marked down for the fact that pumps have genuine compatibility requirements. Quartiles are calculated across the 43 departments with 200+ tenders. Select a row for the breakdown.

DepartmentScoreTenders AuthorisationSampleNamed make Outside rangeReadable
The five parameter families. Authorisation — whether qualifying requires an OEM letter or a proprietary article certificate. Cost to bid — whether a physical sample is required before award. Specification breadth — whether a manufacturer is named without a stated equivalence provision. Threshold proportionality — how far the turnover requirement sits from the median for the same category. Process transparency — response-window length and whether documents are machine-readable. They are conceptually distinct though not statistically independent; buyers using non-catalogue categorisation tend to apply several together.
A methodological trap, and how it is closed. Manufacturer-specific specification cannot be observed inside image-only scans, so a buyer whose documents are all scans would appear to have none and would rank near the top. That would reward lower transparency. The transparency family exists to close the gap: non-readability is scored directly. The diagnostic is the correlation between index and document readability, which comes out at +0.37 — less readable buyers score lower, not higher. Without that family the index would be an artefact of document format and should not be published.

This measures documents, not intent. A fourth-quartile position is a prompt to look, not a finding of any kind against a buyer — a substantial share of narrow specification reflects genuine technical requirement. Departments below 40% document readability are shown, but their specification figure is not reliable and is marked accordingly.

Turnover thresholds sit under a ceiling

Turnover required divided by declared contract value, across the tenders stating both. The distribution terminates at 4.00× — a platform constraint rather than a behavioural pattern. The useful question for a supplier is therefore not whether a threshold is high in absolute terms, but where it sits within the range observed for the same category.

4×+

Median 0.50×. 30.4% of all tenders sit on a round-number rule (0.25 / 0.3 / 0.4 / 0.5 / 1.0 / 2.0 / 3.0 / 4.0), the signature of a formula applied centrally. Yet within the permitted range, a mean 40.2% of tenders sit more than 2× from the median for their own category — the same service, priced for eligibility completely differently by different buyers.

Inside Tender Matcher, this becomes a checkpoint

These findings are not a one-off study. Every one of them is a question that should be asked of every tender before a rupee of EMD is committed — so Tender Matcher runs them as a standing pre-bid gate. Eight checks, answered from the document text, before you decide to bid.

Why a checkpoint and not a report. A vendor does not need to know that 43% of tenders demand OEM authorisation. They need to know that this tender does, that they do not hold that letter, and that the deadline is in nine days — while there is still time to either obtain it or walk away. The value is entirely in the timing.

Bid clinic

If a tender you were counting on turns out to apply three of these parameters at once, the useful next conversation is not with software. Book a working session and we will go through your eligibility profile against these patterns, and how to read a tender you are looking at now. The analysis on this page is historical; the coaching applies it to your situation.

A

Eligibility audit

Your turnover, experience, certifications and OEM authorisations mapped against your live category. Output: the share of your market you are structurally locked out of, and by which lever.

B

Pursuit triage

Your current shortlist scored against category norms — which bars are routine, which are outliers, and where the EMD is likely to be wasted.

C

Representation route

Where a specification looks written for one supplier, what a pre-bid query should actually say, and when it is worth raising at all.

Opens your email app with the message ready — nothing is sent from this page, and nothing you type here is stored, logged or transmitted anywhere. Coaching is general guidance based on published tender documents; it is not legal, financial or bid-writing advice, and no outcome is promised.

How this was measured

The honest version, including what did not work.

Manufacturer detection requires context, not keywords. A manufacturer name only counts when it sits within 200 characters of make / model / OEM / part-number language, and "or equivalent" nearby cancels it. That rule removes 85% of naive brand-keyword hits — "HP" is usually horsepower, "Carrier" is usually a common noun, and a PSU's own name appears on its letterhead. Any published figure without a context rule overstates this roughly sevenfold.
A third of specifications are not machine-readable by anyone. 43,864 attachments (33%) are image-only scans with no text layer — invisible to this analysis and to GeM's own search. Among tenders whose attachments are all machine-readable the rate is 3.06%, not 2.26%. Every rate published here is a lower bound.
Two hypotheses were not supported, and are reported anyway. Short response windows do not co-occur with narrow specification — windows of 10 days or less show a lower rate (1.54%) than longer ones (2.22%). And there is no observable substitution between financial and technical parameters: the rate is flat across every turnover band.
These figures are machine-generated and can be wrong. Every number here comes from automated text extraction and pattern matching over PDF documents — not from human review of each tender. Extraction fails on damaged files, unusual layouts and scans; pattern rules mis-fire on wording they were not designed for. Treat the figures as a well-tested estimate, check anything that matters against the original document on GeM, and tell us if you find an error.

Coverage: tenders closing between 04 Dec 2023 and 08 Sep 2026; 0 are still open, with a further 26,497 in the archive not yet closed and excluded. Figures last regenerated on 08 Sep 2026. Corpus: 73,476 GeM tender folders, 1,83,213 attachment references deduplicated to 1,50,086 unique documents. Extraction via poppler pdftotext in both layout and raw modes. Metadata joins the Jun–Aug 2026 GeM export where available (29,533 tenders). Scan bias verified by OCR on a random sample of 166 image-only documents — no significant difference found. This is a cross-section: it describes structure, not change over time.

Before you use this site. The figures published here are produced by automated analysis of published tender documents and can contain errors. They describe tenders that have already closed — they are not live opportunities. Nothing here is legal, financial or bidding advice, and nothing here is an allegation of wrongdoing against any buyer, official or organisation. Verify anything important against the original document on GeM.

Continuing means you accept our Terms of Use. This site sets no cookies of its own; the Privacy Notice sets out what external services such as the web fonts and the visitor counter can see.