Data Quality Deserts: 20 Industries With Almost No Data

Datadory catalogs 1,744 datasets across 163 industries, and 159 of those industries clear the minimum viability bar and receive an industry brief. Coverage inside those briefs is sharply skewed. Seventy-five of the 159 carry 11 or more primary datasets, while 15 carry zero and another 5 get by on 1-2. That puts 20 cataloged industries — 12.6% of the briefed set — on two or fewer dedicated primary sources.

The zero-primary tier skews consumer-facing: Housewares & Specialties, Wireless Telecommunication Services (data hub), Distillers & Vintners (data hub), Homefurnishing Retail, Tires & Rubber, Airport Services, Consumer Staples Merchandise Retail, Semiconductors, Technology Distributors, Technology Hardware, Storage & Peripherals, Food Retail, Leisure Products, Home Improvement Retail, Regional Banks and Specialized Consumer Services. The housewares-specialties data hub shows the pattern at its clearest: a recognizable consumer category with no native statistical program behind it.

A second tier gets by on a handful of borrowed series. Electric Utilities runs on a single EIA primary (electric-utilities data hub), while Textiles (textiles data hub), Health Care Distributors, Food Distributors and Office Services & Supplies hold two each.

Four industries never made the catalog at all: real-estate-development, real-estate-services, data-center-reits and other-specialized-reits carry no primary or secondary coverage and no brief. Of the 163 cataloged industries, these 4 are true blank spots — there is nothing to pool.

The table below ranks all 20 thinnest briefs by primary dataset count, breaking ties by pooled record count. A primary dataset is one whose first-choice industry is the slice itself; pooled records are tagged in from neighboring industries. Office Services & Supplies shows the ceiling of the workaround — 2 primary datasets but 22 pooled records — while Textiles shows the floor of the second tier, with 2 primaries backed by just 4. Every figure comes from the catalog snapshot dated August 22, 2026.

Three mechanisms recur across the briefs.

First, statistical agencies publish retail demand at NAICS level, not banner level. Housewares & Specialties exists only as an MRTS cut for electronics and appliance stores, and Leisure Products only through County Business Patterns rows such as NAICS 441210 recreational vehicle dealers. The agency never promised a housewares series; it publishes kind-of-business cuts, and the banner you care about is a subset nobody reports separately.

Second, the richest counters in thin verticals are trade associations that publish selectively. The Motorcycle Industry Council releases its Statistical Annuals only on its own schedule, WSTS keeps Logic and Memory billings detail inside its member program, and IFDA's foodservice benchmarks circulate through member channels. Where the association is the only counterparty collecting the data, its own publication rhythm sets the ceiling on what any outsider sees.

Third, some verticals simply lack a public statistical program. Wireless Telecommunication Services rests entirely on the World Bank mobile-cellular subscriptions series (IT.CEL.SETS) (1960-2025) and ITU ICT aggregates (2005-2025). No US or EU agency publishes a carrier-revenue series the way EIA publishes electricity.

Pooling is the catalog's main fix for the deserts. Pooled industry assignments total 2,719 against 1,744 primary datasets — a lift of 975 records, or +55.9% — because almost every record carries secondary industry tags that let a neighboring slice borrow it.

The regional-banks data hub, the best regional-banks datasets ranking and the regional-banks data guide document the same lesson: a desert label describes the primary layer only, not what a buyer can actually assemble.

Only 11 of the 1,124 cataloged sources span 10 or more industries, while 975 are single-industry specialists — so a handful of generalists effectively carry the deserts. Data.gov reaches 60 industries with 76 datasets, the U.S. Census Bureau 46 industries with 64, and Eurostat 39 industries with 52.

The pattern repeats inside the thinnest slices themselves. Across the 20 desert briefs, the U.S. Census Bureau appears in 11 landscape narratives, Eurostat and Data.gov in 10 apiece, and Kaggle in 8. When your own industry page looks bare, the fastest move is to check who already covers your neighbors — one of those five publishers is usually already in the pool.

Buyers substitute what exists for what they wanted. The substitutions are specific, not hand-wavy:

1. Distillers & Vintners reads grape area, yield and production out of FAOSTAT Crops and Livestock Products (244+ countries, 1961-2024) plus Open Food Facts, whose ~4.7 million barcoded products carry ingredients and Nutri-Score. All six pooled records in the slice come from agricultural and product-level sources.

1. Textiles takes cotton and other fibre crops from FAOSTAT Fibre Crops & Cotton Production Statistics: country-year area, yield and production reaching back to 1961.

1. Health Care Distributors leans on the NPPES NPI Registry, CMS's registry of every US National Provider Identifier, shipped as weekly and monthly bulk CSV files.

1. Tires & Rubber queries the NHTSA Vehicle Safety API, whose products array separates Vehicle versus Tire entries down to tire brand, model and size — the whole slice's six pooled records stay inside vehicle-level safety data.

1. **Semiconductors falls back to WSTS's Historical Billings Report (billings from 1986 through June 2026), UN Comtrade Plus HS 8542 trade flows and the World Bank high-tech exports indicator, while product-level Logic and Memory detail remains inside the member program.

1. Airport Services borrows machine-learning snapshots such as the 2015 Flight Delays and Cancellations dataset — 5,819,079 US flight rows — on top of BTS on-time performance series reaching back to October 1987.

1. Wireless Telecommunication Services draws on Hugging Face's 1,012,353-dataset hub, and Home Improvement Retail gets shelf-level detail from the Lowe's Building Materials Product Catalog, distributed as daily, weekly and monthly CSV feeds over SFTP.

Storefront records are the desert's most granular bridge, and they arrive as structured tables rather than page captures. The Office Depot / OfficeMax product catalog carries roughly 49,466 SKU-level rows — price, was-price, brand, model and category placement. Newegg's electronics catalog documents hundreds of thousands of listings field by field, and Fresha's venue directory covers 130,000+ bookable locations with service-level detail.

If you build models or panels (data scientists): check the pooled layer before declaring an industry impossible. Regional Banks went from zero primaries to a working regulator panel — FFIEC Call Reports, FDIC quarterly profiles, HMDA loan records — entirely on borrowed records. The data scientists hub collects the same pattern elsewhere.

If you size markets (market researchers): retail deserts are measurable at NAICS level even when the banner level is closed — MRTS kind-of-business cuts and County Business Patterns rows are the standing substitutes, and the market researchers hub maps them by sector.

If you build prospect lists (sales and growth teams): registries beat benchmarks in the deserts. NPPES covers every US health-care provider identifier, and product catalogs like Lowe's or Office Depot's expose SKU-level assortment the benchmark sellers never publish — see the sales-growth-teams hub for list-ready sources.

The common thread: in a data desert, the question shifts from "where is the industry dataset?" to "which neighboring publisher already measures my proxy?" Start with the pooling table above, then check whether one of the multi-industry giants — Census, Eurostat, Data.gov, Kaggle, Hugging Face — already covers your adjacent slice.