Kind: Measured report
By Asif Hussain, Founder and CEO, SpecIndex
On page 871 of the contract documents for a roof job at Mammoth Cave Hotel in Kentucky, a utilities section finally names a product: one item, listed as basis of design, with the words or approved equal beside it. That page is one of the only hotel spec positions we can point to in the whole index, and the scarcity is the story.
The hotel pipeline looks large from the project count. Read closely, it is a permit pipeline, and it names almost nothing.
Facts at a glance
- We ran search_projects with project_type "hospitality", which returned 3,347 projects in total. (From: Method note.)
- We then ran the same query with announced_within_days set to 660, which returned 2,373 projects, and paged the full window at ten rows per call. (From: Method note.)
- Texas held 917 projects and Florida held 739, together 1,656 of 2,373. (From: Method note.)
- Projects at $5M or more numbered 62 of 2,373. (From: Method note.)
Seven in ten projects in the window sit in Texas and Florida
We counted every hospitality project in the index announced in the last 660 days. About seven in ten of them sit in two states, Texas first and Florida second. The next states, North Carolina, Pennsylvania, and Maryland, trail far behind.
About nineteen in twenty of the window projects are permitting records. This is work being filed and approved, in the main small in scale, rather than work being designed and bid in public.
A permit tells a manufacturer that a building will change. It does not say what the building will buy.
About half the window is banded under one million dollars, and two in five disclose no value at all
The value bands tell the same story as the stage. About half the projects in the window carry a band under one million dollars: renovations, conversions, and small additions.
Two in five carry no value band at all, so the record does not even price them. About one project in forty discloses a value of five million dollars or more.
A rep who adds up this window as hotel construction demand is adding up mostly small jobs and blanks. The large hotel builds that carry full project manuals are a thin slice of the count, and the count does not separate them until you read the bands.
One project in the window holds a document, and it is in Atlanta
Across the whole window, one project holds a document we can open, and it holds exactly one. It is the Centennial Yards Elliott Street Boutique Hotel in Atlanta, a permitting record in the twenty to one hundred million dollar band, announced in July 2026.
One document in more than two thousand projects means the window cannot answer the question a manufacturer actually asks: what do these jobs specify. The project index and the document index are different populations, and for hospitality they barely touch.
The hotel specs we can read belong to public lodgings
The document index does hold hotel and lodge positions. They come from public owners, because public owners publish their manuals.
At Mammoth Cave Hotel, a federal project in Kentucky, the Division 33 section names the specified product "Zurn Z1091" as basis of design, with substitution language that reads "or approved equal" (project "MACA 217837 Mammoth Cave Hotel & Roof Rehab", document "SPECS_2021-06-17_-_MACA_HOTEL_ROOF_REPAIR_-_SIGNED_CONTRACT_DOCUMENTS.pdf", pages 871 to 886). Nine approved manufacturers sit on that section. The position is real and it is contestable in the same breath.
At Multnomah Falls Lodge in Oregon, a Division 26 section names the specified product "Eaton brand Pow-R-Line C switchboard product line" as basis of design, and the substitution language reads "If alternate products are to be used, demonstration of equivalency must be provided in writing." (project "Multnomah Falls Lodge Electrical Upgrades", document "26 24 13-SwitchboardsLowVoltageFeeders-20200505.pdf", pages 1 to 7). Five approved manufacturers sit on that section, and any rival has to prove equivalence on paper before price enters the room.
A park hotel and a waterfall lodge are not the private hotel market. They are the part of the market whose paper is public, and they are what a spec index can honestly show today.
A permit count is not a spec pipeline
Our judgment: in hospitality, the project count and the spec record describe two different markets, and mixing them produces bad targeting. The counted pipeline is Texas and Florida permit work, small in value where it is valued at all, with no manuals behind it.
The readable spec positions are public lodging jobs, scattered across park and lodge projects, where the named product and the openness wording can be checked line by line. A manufacturer who aims a spec program at the permit count is aiming at records that cannot name his product or his rival.
What to do with this
If you sell into hotels, read the Texas and Florida flow as weather, not as a chase list: it says small hotel work is being permitted at volume in two states, and nothing more. Build the chase list from documents instead, and expect it to be public work at first: park hotels, lodges, and municipal projects whose manuals are posted.
Test any hotel pipeline claim, including ours, with one question: can you open the manual. In this window, one project in 2,373 has a manual to open. If a count cannot show its manuals, treat it as a permit count and price it that way.
If you write specs for public lodging, know that your manual is the visible record for the whole sector. The openness line you choose, or approved equal against a written proof of equivalency, is the competition rule every bidder will work under.
Method note
Population and queries, as of October 16, 2026. We ran search_projects with project_type "hospitality", which returned 3,347 projects in total. We then ran the same query with announced_within_days set to 660, which returned 2,373 projects, and paged the full window at ten rows per call.
State counts, stage, and documents were taken from the location, status, and documents_held fields on each row. Texas held 917 projects and Florida held 739, together 1,656 of 2,373.
Permitting records were 2,270 of 2,373. One project in the window holds documents, with one document in total.
Value bands were counted from the value_band field, with records carrying no band counted as undisclosed. The distribution was: under $1M, 1,232 projects (51.9 percent); undisclosed, 951 (40.1 percent); $1M to $5M, 128 (5.4 percent); $5M to $20M, 38 (1.6 percent); $20M to $100M, 16 (0.7 percent); $100M and above, 8 (0.3 percent). Projects at $5M or more numbered 62 of 2,373.
The exhibits came from find_brand_positions, a separate document index and a different population from the project index above. Returned position totals on October 16, 2026 were 1,995 for Armstrong, 1,076 for Zurn, 1,503 for Eaton, and 207 for Moen. We paged the returns at five rows per call, scanning all 207 Moen returns, the first 850 Zurn returns, the first 800 Armstrong returns, and the first 375 Eaton returns, and filtered them client side to project names containing Hotel, Resort, Inn, Lodge, Hospitality, or Casino.
That substring filter returns false positives, including Minneapolis and Cincinnati on the Inn term and the town of Lodge Grass in Montana, which we discarded by reading each matched name. The genuine matches in the scanned returns were public lodging projects, among them the two exhibits, a second Timberline Lodge project in Oregon, and a lodge and cabins project in Missouri.
Both exhibits were announced in 2021, outside the 660 day window. We quote the specified_product and substitution_language fields exactly as returned, with the project name, document_title, and page span exactly as returned.
Limits, stated plainly. The project index and the document index are different populations with different coverage, and neither is a census of hotel construction.
Private hotel projects bid without public documents do not produce a readable manual, so the document index understates private hospitality work by construction. The window count describes permits and announcements the index holds, weighted to the states its sources cover, and the value distribution rests on bands the source records carry, which two in five records do not carry at all.