Methodology

What the Awards Are

The Vedex Index Awards rank alternative-data vendors within tightly defined categories — a topic like Satellite Imagery, or a predicted metric like Best Data for Predicting Same-Store Sales. The goal is simple: help a buyer see, at a glance, which vendors are genuinely strong for a specific use case, and why.

Every award is the output of a disclosed, rules-based method. There is no secret sauce and no editor putting a thumb on the scale. Below is exactly how a vendor gets from raw data to a band and a rank.

Who Is Ranked: Data Vendors Only

The Vedex index lists every company we find on the data marketplaces we track — and those marketplaces also list software, developer tools, ETL products and consultancies. Those companies keep their Vedex profile and score, but they are not ranked on any leaderboard, in the rankings export, or in the Awards. A buyer comparing data vendors should not find a DevOps platform or an ETL tool among them.

The same rule runs everywhere a vendor is ranked. In order:

1. No score, no rank. A vendor without a Vedex score is never ranked.
2. Categories are read, not trusted blindly. Each of a vendor’s categories is classified as software / services (e.g. DevOps, CI/CD, ETL, data integration, business intelligence, implementation, managed services, consulting), as data (e.g. “… Data”, satellite imagery, foot traffic, sentiment, ESG, firmographics, market data), or neutral.
3. Data-selling vendor types are ranked — data provider, aggregator, index provider, exchange, marketplace — unless every classified category is software.
4. Tools and consultancies are not ranked — vendor type tool / service or consultancy — unless their categories name a data product and data categories outnumber software ones (vendor types are sometimes mislabelled).
5. Everyone else (no vendor type, “platform”, nonprofit) is ranked when data categories outnumber software ones, and not ranked when software categories are at least as many. A vendor with no categories at all is ranked only if it is listed on a data-only marketplace (Datarade, Snowflake Marketplace, Nasdaq Data Link, Exabel); otherwise it is marked unclassified and not ranked.
6. A short, published correction list overrides the rule where it is wrong. Every entry is below, with its reason.

Eligibility never changes a score. In the Awards, a placement held by a vendor the rule identifies as software, tooling or a consultancy is removed and the remaining placements are renumbered (bands and scores unchanged); each removal is listed on the award page. Vendors that are only unclassified keep their Award placement, because the Awards already place vendors from their products’ topic tags. The live counts of ranked and excluded vendors are shown on the rankings page. Rule version 2026-09-25.

Like-for-Like: Category Groups

A single leaderboard that puts a news-analytics vendor next to a market-data terminal compares things buyers never choose between. So the rankings page opens on category leaderboards: each vendor is ranked only against vendors in the same category group, and each board says what it is compared within. The single cross-category board is still available, and labelled as such.

Each of a vendor’s categories maps to the first group below whose keywords it matches, so “Financial News” is News, not Market Data, and “Credit Card Data” is Consumer Transactions, not Credit. Generic categories that say nothing about the data itself (e.g. “Market Analysis”, “Predictive Analytics”) and software categories map to no group. A vendor sits in the group holding the largest share of its own categories, and also in a second group if that group holds at least half as large a share — never more than two. Only if none of its own categories map to a group are its products’ categories used instead.

1. News, Sentiment & NLP — News analytics, sentiment, social and text-derived signals.
2. Web, App & Digital Activity — Web traffic, clickstream, app usage, search and scraped web data.
3. Consumer Transactions & Commerce — Card and receipt panels, POS, e-commerce and retail activity.
4. ESG, Climate & Weather — ESG scores, emissions, environmental and weather data.
5. Geospatial & Location — Satellite imagery, foot traffic, POI, mobility and mapping data.
6. Energy & Commodities — Energy, commodity prices, metals, mining and agriculture.
7. Supply Chain, Trade & Logistics — Trade flows, shipping, freight, supplier and logistics data.
8. Healthcare & Life Sciences — Claims, clinical, real-world and pharmaceutical data.
9. Real Estate & Property — Property records, valuations, construction and housing data.
10. Risk, Credit & Compliance — Credit ratings, KYC/AML, sanctions, fraud and legal data.
11. Private Markets & Funds — Private companies, PE/VC, fund and manager data.
12. Company, B2B & Firmographics — Firmographics, contacts, intent, technographics and company data.
13. Audience, Identity & Marketing — Audiences, identity graphs, advertising and marketing data.
14. Automotive, Travel & Leisure — Vehicles, travel, hospitality, sports, gaming and entertainment.
15. Macro, Economic & Demographic — Economic indicators, census, population and public-sector data.
16. AI Training & Synthetic Data — Training, labelled and synthetic datasets for ML.
17. Market, Pricing & Reference Data — Market, pricing, reference, index and fundamental data.

A group is shown once it has at least three ranked vendors. Free-text marketplace categories are messy, so the grouping is rule-based rather than perfect; if a vendor sits in the wrong group, tell us and we will correct it.

The Pillars

Each vendor is scored on seven pillars, each on a 0–100 scale. The pillars are the shared vocabulary across every category — so a Band 1 in one category is built from the same measured dimensions as a Band 1 in another.

The Composite: A Geometric Mean

The seven pillars are combined into a single composite score using a geometric mean, not a simple average. This is a deliberate choice.

Percentile Banding & the Vedex Star

Raw composite scores are not directly comparable across categories of different size and competitiveness. So ranking is done by neighborhood percentile — where a vendor sits relative to its qualifying peers in the same category. Percentiles map to bands:

Bands describe standing within a peer group. They are not absolute grades — a Band 1 in a small category is a claim about that category, not about the whole market.

The Coverage Gate & Confidence Flags

We only rank vendors we actually know enough about. Each vendor carries a coverage ratio — how much of the scored surface we have verified data for. A vendor must clear a minimum coverage gate to qualify for an award at all. Below it, we would be guessing, and we would rather say nothing.

Above the gate, every award carries a confidence flag — high, medium, or low — that signals how complete the underlying data is. A medium- or low-confidence award is a real result, but it comes with an honest caveat that some inputs are thinner than we would like. Confidence is about how much we know, not how good the vendor is.

Computed First, Editorial Narrates It

The rank is computed. Full stop. The pipeline calculates pillar scores, the geometric-mean composite, the neighborhood percentile, the band, and the star — and only then does an editorial layer write the headline, the rationale, the strengths, and the “watch” note.

Recusal for Affiliation

Vedex is not at arm’s length from every vendor in its index. Where it is not, the vendor is on our assessor-affiliation register and is recused from the Awards: it is dropped before category membership is computed, so it cannot qualify, cannot be banded, cannot take a Vedex Star, and cannot appear in a quadrant. The recusal list is published as part of every edition, so the exclusion is visible rather than silent.

Recusal is an Awards rule, not an index rule. A recused vendor keeps its published index score and its position in every Vedex ranking; it is scored by this same methodology, from the same sources, with no adjustment, and it is marked Affiliated wherever it appears. Full disclosure on the independence page.

A Crucial Caveat: Evidence, Not Alpha

For predicted-metric categories — the “Best Data for Predicting X” awards — the causal and evidence scores reflect research-backed relevance, not measured trading alpha.

Disclaimer

For how we keep the ranking honest and free of pay-to-play influence, see our Editorial Independence & Monetization Policy.