Methodology: where the data comes from and how names were matched

Every figure on this site is copied from one of four public datasets built on 4 August 2026 from government documents: the GujRERA portal, DICDL's annual reports for FY2023-24 and FY2024-25, the Gazette of India SEZ notification of 9 April 2026, and DICDL's land pricing page. Nothing is estimated, interpolated or carried over from memory. Where a source is silent this site prints no data.

Name matching across registers is the weak point of any aggregation like this, so the rule is deliberately strict and it is applied by code, not by judgement.

1. Normalise. Convert to upper case, remove a leading M/s or Messrs, and reduce every character that is not a letter or a digit to a single space. The result is the identity key.
2. Merge only on an exact key. Two records become one company only when their identity keys are character-for-character identical. No legal form is stripped before this test, so ABC DEVELOPERS and ABC DEVELOPERS PRIVATE LIMITED produce different keys and are never merged.
3. Or merge on an alias the source itself declares. The only other route. The FY2024-25 annual report prints M/s Tata Semiconductor Manufacturing Private Limited (formerly M/s Semifab Private Limited), so the Semifab record from the FY2023-24 report is attached to that company and the declaration is shown on its page.
4. Split joint ventures. A contractor field naming a venture, such as the joint venture of M/s SCC Infrastructure and M/s P. R. Patel and Co., is split into its named members and the record is attached to each with a joint-venture flag. The award value is stated once for the venture, so it is never divided between members and never added into a member's sole-name total.
5. Flag, never merge, the near misses. Where two entries share a core name once the legal form is removed, both entries stay in the register and each is tagged UNCONFIRMED PAIR with a link to the other.

The false-match risk, stated plainly

This rule produces two kinds of error and it is worth being clear about both. It will split one company into two entries whenever a register prints its name without the legal suffix, which is common. It will merge two genuinely different companies whenever they happen to be registered under exactly the same name string, which is rare but possible. None of these sources publishes a company identification number, a registered office or an incorporation date, so there is no field available that would settle either case. The conservative choice is to split rather than merge, because a wrongly split entry understates a footprint and can be corrected by anyone with a document, whereas a wrongly merged entry attributes one company's records to another.

2 unconfirmed pairs in a set of 39 companies · source GujRERA and DICDL annual report FY2024-25 · as of 4 August 2026

5 name families · no relationship between them is asserted · as of 4 August 2026

Two records appear in both the land dataset and the tender dataset, because DICDL's annual report describes them in both ways: a plot awarded by e-auction to M/s GAP Associates with a letter of award dated 14 November 2024, and a plot awarded by e-auction to M/s Seksaria Buildcon LLP in joint venture with The Amore Banquet, letter of award dated 24 December 2024. The tender dataset itself separates them, counting two of its eleven awarded records as land e-auction letters of award rather than works contracts.

This site follows that separation. Both are counted once, in the land register, with both citations shown. Neither is counted as a works contract, and neither is counted twice towards a cross-register footprint. Without this reconciliation Seksaria Buildcon LLP would appear to sit in all three registers, which would be wrong.

These are gaps in the underlying government documents, not gaps in the extraction. They are reproduced here because a register that hides its holes is worse than no register.

Land

1. No per-company area for the four named industrial allottees. The FY2024-25 annual report states that 1,519 acres of industrial land is available and 530 acres already allotted, at 31 March 2025, and names Tata Semiconductor Manufacturing, Renew Photovoltaics, Polycab India and Inox Air Products against that combined figure without splitting it. The FY2023-24 report gives 522 acres allotted at 31 March 2024. This site does not compute a difference between those two, because the source does not present one as a year's allotment.
2. The list is explicitly not exhaustive. It ends with the words other prominent industrial proponents, which is recorded here as an unnamed entry and is not a company.
3. No allotment date on any allottee record except the two e-auction letters of award.
4. No consideration, anywhere. The published rate card is undated and is an asking rate. The audited accounts record a nominal annual rent. Neither is evidence of a price paid.
5. One unit inconsistency inside a single document is left unreconciled: DSIRDA's equity land is stated as 4,830.83 hectares in the audited note and as 48.31 square kilometres in the Board's Report of the same annual report. Both are recorded as printed and no conversion has been performed.
6. The 66.166 hectares in the SEZ gazette is the notified zone footprint, not an allotment deed. It states no allotment date, no tenure, no lease term and no consideration, and it is not necessarily the total land that allottee holds in Dholera SIR.

Contracts

1. Tender identifier is null on every record in every source, so no award can be joined to a tender notice.
2. Estimated contract value is null on every record, so bid against estimate cannot be computed.
3. Completion period is null on all ten awarded contracts drawn from government sources. The annual report gives construction start dates but never a contract duration.
4. For six of the construction awards the annual report states a total project cost rather than an explicitly labelled contract amount. Only the administration and business centre award is printed as a contract amount, 72,31,11,111 rupees.
5. No award value at all is disclosed for the terminated water treatment plant contract, in either annual report. It is left as no data rather than back-solved from the bank guarantee amounts published alongside it.
6. The procurement portal returned nothing. The awarded-tender API returned zero rows for DICDL and also zero rows for all clients, which is evidence that the capture failed rather than evidence that no awards exist.
7. One widely repeated contract claim was deliberately excluded because no primary source states a value, a date or a client for it. It appears only on property marketing sites. Commercial tender aggregator sites were not used as sources at all.
8. One record, the service area building award to Nila Infrastructures Limited at INR 46.21 crore, is tagged REPORTED. It comes from trade press and is not corroborated by any government document opened, its award date is unknown, and its scope description conflicts with DICDL's own administration and business centre award. It is kept, tagged, and never mixed into an OFFICIAL total without the tag being visible.

RERA

1. Project land area is blank on 10 of the 33 registrations, because the government page renders that line blank. Those rows print no data.
2. Three registrations carry the evidence tier INFERRED rather than OFFICIAL, meaning the Dholera location rests on the registration number and village rather than on the approving authority field.
3. Quarterly progress filings are counted as filed against scheduled, as the portal reports them. A shortfall can have causes the portal does not publish, so this site publishes the two counts and does not rank on them or characterise them.
4. Complaint counts are shown because they are public. They are not ranked and they are not described, because a complaint is an allegation and its outcome is not published in a form this dataset captured.

The company set itself

1. None of these registers publishes a company identifier, a registered office or an incorporation date, so entries are matched on name alone.
2. The register covers only what these four datasets contain. A company holding land or a contract in Dholera SIR that is not named in any of them is absent, and its absence means nothing about it.
3. Addresses and names are printed as the register prints them, including the register's own spelling. One typographic change is made: an en dash inside a printed address is rendered as a hyphen.

Write to corrections@dholera-group.com. State the company, the page, the figure you say is wrong and the public document that shows the correct value. Corrections that cite a public document are applied and the change is dated here. Corrections that ask for a true record to be removed are not applied, and neither are requests to add a company that no public document names.

Corrections applied to date: none. This is the first edition, published 4 August 2026.

Compiled and published by Bhavik Sarkhedi, ORCID 0009-0005-8094-0688. This register has no commercial relationship with any company listed on it, sells nothing, accepts no listing fee and takes no advertising. It is a register, not a broker. Nothing on it is advice about buying, selling or investing in anything.