Method and limitations

Why there are seven lists and not ten

This site publishes seven ranked lists because seven is how many genuinely different columns this dataset holds. Each one sorts on a real field, or ranks a different entity type, and the rule for every one of them is printed on its own page. Rewording a ranking to make an eighth page would be duplicate spam, and it would also be a lie about how much evidence exists.

Four further axes were tested against the data and rejected. They are named here rather than quietly dropped, because a reader deserves to know what was considered:

• By intended use or sector, at parcel level. Rejected: exactly two of the twelve parcels state an intended use, a mixed use development and a hotel. A two row list ranked on a field eighty three per cent of the register leaves empty is not a list. The sector question is answered instead by the rate card list, which ranks eight real land use categories on a real published column.
• Government versus private allottees. Rejected as a page: honestly built it is three rows long, one government holder, six private holders and one placeholder the source declines to name. It is a column on the allottee list and a counted section underneath it instead.
• Anchor industrial parcels. Rejected: this is a filter of the register rather than a new column, and the four named industrial allottees already carry a section of their own on the allottee list. Publishing the same four rows again under a new heading is the exact duplication this register exists to avoid.
• By land value or consideration paid. Rejected on the strongest ground of all: no document in this source set states a price paid for any named parcel. The two money figures attached to the e-auction awards, approximately Rs 2,000 crore and approximately Rs 35 crore, are stated total project investment and not land consideration, and ranking anything by them would misdescribe both awards. That axis will exist here when a registered deed does, and not before.

The same discipline governs the charges page. Fees, deposits, transfer charges and penalties are published as a reference table and not as a ranking, because a fee in rupees, a percentage of market value and a penalty rate per month do not share a scale.

Where the data comes from

The unit of record here is the parcel: who holds land in Dholera Special Investment Region, how much of it, where, since when and on what terms. Six documents were read for it. Four are official PDFs, one is an archived capture of an official portal page, and one turned out to be irrelevant and is logged anyway. The full list, with what each one produced, is on the sources page.

How it was extracted

Each PDF was opened and read in full rather than searched for keywords. Every record carries the file it came from, the PDF page number, a human readable page label such as the printed page number or the form and clause reference, and a verbatim quote of the sentence the figure sits in. Those four fields travel with the record into the JSON and the CSV, so any figure on this site can be checked against its source in one step. Where a source used a character that does not extract cleanly, such as the rupee sign in the annual report, the extraction marks it in the quote rather than guessing.

How to verify a row, and where that gets hard

Every row on every list carries its evidence in the row itself: the document, the page, the record reference and the date the row was last verified. Clicking the citation opens that record’s own page, which prints the verbatim sentence the figure sits in, so a reader can check the extraction without opening a PDF at all.

One source can be opened directly on a government server and is linked as such wherever it appears: the DICDL Land Pricing page at dholera.gujarat.gov.in/land_pricing, which is the canonical URL printed on the page itself. The other four documents are a declared gap. The dataset records a full citation for each of them, gazette number, notification number, file number, report year and page, but it records no permanent public URL, so none is printed here. Inventing a plausible looking government link would be exactly the kind of unverified detail this register exists to refuse. If you hold a permanent official URL for the gazette notification, the DICDL annual report, the allotment policy or the cleanroom procurement, send it and it will be added.

What is deliberately not done

• No number is estimated, interpolated or rounded to look tidy.
• No unit is silently converted. The one conversion on the site, square metres to hectares on the area ranking, is stated on that page with its exact definition and the printed figure is shown next to it.
• No row is dropped because it is empty. A parcel with no area still appears in the area ranking, marked UNKNOWN and listed under the ranked ones.
• No inconsistency in a source is reconciled. Where the annual report states the same land as 4830.83 hectare in one place and 48.31 sq. km in another, both are recorded and neither is corrected.
• Nothing is inferred from policy. The policy default of a 99 year lease is recorded as a policy term and is never attached to a parcel whose own document does not state a tenure.

What is missing, stated as limitations

These are the declared gaps in the dataset, published in full rather than summarised. They are the honest boundary of what this register can answer.

1. NO PER-COMPANY AREAS. dicdl_ar25.pdf p.20 names four allottees (Tata Semiconductor Manufacturing, Renew Photovoltaics, Polycab India, Inox Air Products) but gives only the 530-acre aggregate. It states no area, plot number, survey number, allotment date, lease term or consideration for any individual company. parcel_area is null for all four.
2. ALLOTMENT DATE IS MISSING ON MOST PARCELS, BUT NOT ALL. Corrected 2026-08-04 after adversarial verification refuted the earlier blanket claim. dicdl_ar25.pdf states a Letter of Award date for two e-auction awards: M/s GAP Associates, 14 November 2024 (p.28), and M/s Seksaria Buildcon LLP in JV with The Amore Banquet, 24 December 2024 (p.29). Both are now records in this dataset. For every OTHER named allottee, including the four on p.20 and the Tata SEZ footprint, no document in this source set states an allotment date and allotment_date remains null.
3. NO CONSIDERATION PER PARCEL. No document states what any named allottee actually paid. The rate card gives the published asking rate; the audited accounts give the Re. 1/sq.m nominal annual rent. Neither is evidence of a specific transaction price.
4. THE ALLOTMENT POLICY PDF CONTAINS NO RATE CARD. land_allotment_policy.pdf was read in full (99 PDF pages, 27 policy pages plus annexures). It defines 'Reserve Price' as a price the Land Management Committee fixes from time to time and leaves every price field in the allotment letter and lease deed forms blank ('Price of land per sq. mt Rs. ____'). It publishes no per-square-metre figure. The rate card records in this dataset therefore come from the DICDL portal Land Pricing page (lp.html, canonical https://dholera.gujarat.gov.in/land_pricing), which is a separate official source.
5. THE RATE CARD IS UNDATED. Its table is headed 'Revised Land Allotment Rate' but the page does not say when it was revised, from what, or with effect from when. The only date on the page is a site-wide footer 'Last Updated 07/07/2026' that cannot be tied to this table. value_date is null on all eight rate records.
6. gil2019.pdf CONTAINS NOTHING FOR THIS DATASET. It was opened and read in full (22 pages, 119,714 characters). It is a Gujarat school laptop RAM-upgrade list, district / taluka / village / UDISE code / school name / number of laptops, 1,609 rows totalling 3,173 laptops. Zero occurrences of allot, lease, plot, survey, hectare or acre. It contributed no records.
7. nicdc_cleanroom.pdf HAS NO ALLOTMENT. It yielded one planned parcel only: a 2,000 sq.m 'Site Development Area', expressly labelled a 'Preliminary Value', for a Common Cleanroom Facility to be built on PPP basis. No allottee has been selected; the document is a November 2025 RfQ cum RfP to hire the design and bid-management consultant. Tagged OFFICIAL_TARGET, not an allotment.
8. NO SURVEY-NUMBER LEVEL PARCEL REGISTER EXISTS IN THIS SOURCE SET. Only one parcel in the whole set carries a plot and survey reference: the Tata SEZ footprint (Plot 255 and 238 part-2, part of Survey No. 151, Dholera village, TP Scheme 2A). Every other holding is described only as 'in Dholera SIR' or 'in the Activation Area'. A true parcel register would need TP scheme final plot registers or registered index-2 records, which are not in this set.
9. ROAD LAND HAS NO ABSOLUTE AREA. The 99-year, Rs 1 consideration lease-back of road land from DSIRDA to DICDL is stated only as '15% of land appropriated for roads'. No hectare or acre figure is given, so parcel_area is null.
10. UNIT INCONSISTENCY INSIDE ONE DOCUMENT, LEFT UNRECONCILED. dicdl_ar25.pdf states DSIRDA's equity land as '4830.83 hectare' in the audited note (p.71) and as '48.31 sq. km ... valued at Rs 2,898.50 crore' in the Board's Report (p.20). Both are recorded verbatim. No conversion was performed, per the no-tidying rule.
11. THE 1,519 / 530 ACRE WORDING IS AMBIGUOUS IN THE SOURCE. The Board's Report says 1,519 acres 'is now readily available for allotment' and then that 530 acres 'have already been allotted' out of it. Reproduced as printed. Do not publish a derived 'still available' figure from these two numbers.
12. NO TENURE STATED ON THE FAB PARCEL. The SEZ gazette establishes the notified 66.166 ha footprint but is silent on whether the land is held on lease or on sale, on the lease period, and on consideration. Those three fields are null. The policy default (99-year lease) is recorded separately as a policy term and must not be silently attached to this parcel as if the gazette said it.

One further limitation, restated in plain terms

All five input PDFs were also searched for the Vedanta and Foxconn names, for process node strings including 90nm, for a December 2026 date, for the phrase first silicon, and for fab civil progress percentages. There are zero substantive hits. This dataset therefore carries no process node claim, no production date claim and no construction progress claim of any kind, and neither does this site. What this register does refuse to repeat is set out on the accuracy page.

How to correct this register

Send the document. A correction is accepted when it names a public source, gives its page or clause, and quotes the sentence that carries the figure. Corrections that arrive that way are applied to the dataset, logged on the change log with the date and what changed, and the edition stamp in the masthead moves. Corrections that arrive without a source are not applied, however confident they sound. Write to [email protected].

Design and independence

This is a register, not a listing. Nothing on it is for sale, no developer or broker pays to appear or to be ranked, and there is no enquiry form anywhere on the site. There is no photography either, and no parcel plan is drawn, because the source set publishes no parcel geometry and drawing one would mean inventing it.