Datarade gives you a long list of providers and, for each, a listing written by the provider. This guide turns the listing into a scoring sheet of six rows: coverage, freshness, sourcing, delivery, licence and a sample test. Each row names the field of a Datarade listing to read, the question to ask and the check to run on a sample, and the scores add up to one number per provider that a procurement or compliance team can review. How Datarade itself works is in the guide for buyers. The statements about Datarade come from its listing pages, publishing policies and provider help centre, read on 4 October 2026.
The sheet
Datarade's company data page lists the factors its editors consider, among them accuracy, coverage, timeliness, historical depth, formats, delivery methods and pricing models. The sheet groups them into six rows and adds the checks only you can run.
| Row | Read on the listing | Ask the provider | Check on the sample |
|---|---|---|---|
| Coverage | Geographic coverage, data volume, attributes | Which of the companies I track are in it, matched on which key? | Match rate of the companies you track on one key |
| Freshness | Frequency of updates, temporal coverage | When was each fact last observed, and is history rewritten? | Age of the last-update field, and change between two deliveries |
| Sourcing | Source, and the profile's account of how data is sourced and collected | A document for each source type and collection method | Twenty records checked against the companies' own websites |
| Delivery | Delivery methods, formats | The schema, the file names, the schedule and how to pull an increment | Load it with your own pipeline |
| Licence | Terms of use, pricing models | Grant, term, deletion, redistribution, and the price for the companies you track | Read before you test |
| Sample test | The sample preview | A sample for the companies you track in the delivery format | Empty values, duplicates and join keys |
Score each row from 0 to 3. A 3 means the evidence is documented and passes your check, a 2 means it passes with a gap you can work around, a 1 means it is stated and not shown, and a 0 means it fails or is refused. Weight the rows to your use, and treat a 0 on delivery or licence as a stop whatever the total.
Coverage
A provider's record count says nothing about your list, so the denominator is your own list. Take a sample for the companies you track, join it on one stable key and count. A website domain or an identifier such as an ISIN beats a company name, which matches differently under every spelling. The Attributes filter in Datarade's product search lets you start from products that list the key you need.
-- Your companies, one row each, against the provider's sample
select count(*) as total,
count(s.domain) as matched,
round(100.0 * count(s.domain) / count(*), 1) as match_rate_pct
from my_companies u
left join (select distinct domain from provider_sample) s
on s.domain = u.domain;The distinct stops duplicate keys in the sample from inflating the match. A match on the key is not coverage of the fields: the quality query below counts empty values in the fields you need.
Freshness
The listing gives a frequency of updates and a temporal coverage, which Datarade's policies define as recency and historical lookback. Both are claims about the dataset. The sample shows the age of each record. Ask for the timestamp that says when a fact was last observed, not when the row was last written, and ask whether past periods are rewritten. A provider that rewrites history changes reports you have already sent, and a back-test on rewritten history uses information nobody had on the day. A refresh can run daily while a given fact is months old, so the cadence on the listing and the age of a fact are separate measures.
-- Share of records last updated inside the interval the listing claims (30 days here)
select round(100.0 * avg(case when last_updated >= current_date - interval 30 day
then 1 else 0 end), 1) as updated_in_interval_pct,
min(last_updated) as oldest
from provider_sample;Then compare two deliveries a week apart. Rows for closed periods should not change between them.
Sourcing
The listing states a source, and a profile must explain how the provider sources and collects its data. The policies say a listing is removed when the documentation that proves how personal data or web data is handled is inadequate. Score on documents received, not on statements. Ask for a document for each source type that gives the collection method, says whether the collector identifies itself and obeys robots.txt, and says whether personal data is included and on what basis. The sourcing statement of Fokals shows the shape of an answer. Then check twenty records against the company's own website or filing. A fact you cannot find at its stated source scores 0 for that record, and a provider that cannot name a source scores 0 for the row.
Delivery
Datarade's request form lists the ways a buyer can ask for data to be delivered, among them S3 bucket, SFTP, REST API, Snowflake share, Databricks Delta Share, Google BigQuery, Azure Blob Storage, Google Cloud Storage, email, feed APIs and streaming APIs. Score delivery against your pipeline. A 3 means the data arrives in a form the pipeline already takes, and a 0 means a build nobody has budgeted. Ask for the schema, the file naming, the refresh schedule and whether an increment can be pulled from a bookmark, and load the sample with your own code, not by hand. The comparison of API and bulk files sets out when each pull fits.
Licence
Datarade says it is not a party to contracts between customers, so the licence is the provider's. Each listing links its terms of use, which the policies say detail the licence grant and its restrictions. Read them for the use you intend: internal analysis, embedding in a product or redistribution. Read the term and what you delete when it ends. Note whether the price is one-time, ongoing or usage based, the three options on the request form. The policies list reselling data as a standalone product among the uses a provider may not advertise, so ask for any redistribution right in writing. The guide to what a data licence covers lists the points to settle.
The sample test
A preview on a listing is the provider's choice of rows. The help centre says a CSV preview shows the first ten, and the policies require a sample to be representative of the product's quality and completeness and to hold the product's attributes. Treat the preview as a first look, and ask for a sample for the companies you track in the delivery format. Then run the queries here and the two above. They use the dialect of DuckDB, and the guide to exploring bulk company data with DuckDB shows how to read the files.
-- Duplicates and empty values on the fields you need
select count(*) as row_count,
count(distinct domain) as distinct_keys,
round(100.0 * avg(case when employees is null then 1 else 0 end), 1) as employees_empty_pct,
round(100.0 * avg(case when industry is null then 1 else 0 end), 1) as industry_empty_pct
from provider_sample;A row count above the number of distinct keys means one company appears more than once, and a join will multiply its rows. Keep the sample and the answers with their date, because a compliance review may ask for both.
Datarade labels the Quality block that some product pages show as self-reported by the provider. A figure there is the provider's own measure. The policies bar a listing from misrepresenting data quality, and a test on the companies you track is how you confirm the figure that matters to you.
Scoring, with an example
Acme Robotics (illustrative) weights the rows: coverage 30, freshness 20, sourcing 15, delivery 15, licence 10 and sample test 10. The three providers are invented for the example, and the total is the weighted score as a percentage of the maximum of 300.
| Provider | Coverage | Freshness | Sourcing | Delivery | Licence | Sample test | Total | Result |
|---|---|---|---|---|---|---|---|---|
| A | 2 | 3 | 2 | 3 | 3 | 2 | 81.7 | Shortlisted |
| B | 3 | 2 | 3 | 0 | 2 | 3 | 75.0 | Stopped by delivery |
| C | 1 | 2 | 3 | 2 | 3 | 1 | 61.7 | Third |
Provider B would outrank C on the total, and its 0 on delivery stops it. The scores are illustrative, not findings about any provider.
How Fokals reads on the sheet
Fokals is delivered direct, by REST API and as bulk files, which you load with your own pipeline. The sheet applies to Fokals as to any provider, and these are the answers from its own documents.
| Row | What Fokals states |
|---|---|
| Coverage | Listed companies worldwide (equity listings in 79 countries), the brands they own and verified private companies, each with one stable Fokals company ID, and ticker, MIC, ISIN, LEI and FIGI on every listed company |
| Freshness | Refreshed daily to weekly by dataset; daily and weekly tables are written once after the period closes and never revised; every record carries the time it was observed, so the data is point-in-time by construction |
| Sourcing | Sourced from first-party company sources and public records, and processed in-house; company-level data throughout, described in the sourcing statement |
| Delivery | A REST API of 25 endpoints (JSON, one bearer key per client, cursor pagination) and bulk exports as JSON, JSON Lines or CSV with a manifest naming the period, label versions and licence |
| Licence | By written agreement for internal use, embedding in a product or redistribution |
| Sample | Sent on request for the companies you track, with the data dictionary and methodology |
A pipeline that pulls from an API or loads files scores Fokals high on delivery. The delivery page describes what the API and the exports hold.
Frequently asked questions
How do you compare data providers on Datarade?
Use one sheet for every candidate. Read the same listing fields for each, then score coverage, freshness, sourcing, delivery, licence and a sample test from 0 to 3, weight the rows to your use, and treat a 0 on delivery or licence as a stop. Base the coverage and sample scores on a sample for the companies you track, not on the preview shown on the listing.
How do you test a data provider's sample before buying?
Ask for a sample for the companies you track in the format you would receive. Join it to your list on one stable key and count the matches, then measure empty values and duplicate keys on the fields you need, and check the age of the last-update field against the claimed refresh. Check twenty records against the companies' own websites, and load the sample with the code you would use in production.
How do you measure the coverage of a company dataset?
Divide the number of companies found in the provider's sample by the number you asked about. Use one stable key such as a website domain or an ISIN, count each company once, and then measure how many of the matched rows hold the fields you need. The provider's own record count says nothing about your list.
How do you check that a provider's data is fresh?
Compare the claim on the listing, a frequency of updates, with the sample. Take the field that records when each fact was last observed and count how many records fall inside the claimed interval. Then ask for two deliveries a week apart and check that rows for closed periods have not changed. A cadence of refresh and the age of a fact are different measures.
Does Datarade verify the claims on a product listing?
Datarade's publishing policies say a listing may not advertise attributes the product lacks or misrepresent data quality, and that the provider is responsible for following them. Its data protection team verifies certifications before they are published, and the Quality block that some pages show is labelled self-reported by the provider. A test on a sample for the companies you track is how you confirm the claims that matter to you.
The queries and code on this page are examples to adapt. Test them in your own environment before you rely on them.
What this page says about the products it names was checked against their public documentation on 4 October 2026. Product and company names are trademarks of their owners. Fokals is not affiliated with them or endorsed by them.