Data providers are companies whose product is the data itself. They collect it from many sources, clean it, match it every day and sell it, usually by subscription. They are also called data vendors, and their business model is called data as a service (DaaS). A company buys their data when it needs to know about the world outside its own systems.
Consider a company in Rajkot that makes air coolers and sells them across India. Its own data warehouse tells it exactly what it sold, to whom and where. But it cannot see its rivals' sales, so it cannot tell whether it is gaining or losing customers. It cannot see which brands are importing coolers and compressors into India. And it has not read everything that a listed rival told its investors. These answers come from outside, and the company has to buy them.
The first group of providers measures what people buy and watch. Retail measurement collects sales from shops' tills and from regular counts at a sample of shops, and tells each brand its share of the market. Audience measurement counts how many people watched each TV programme, the figures on which TV advertising is priced. Market research data estimates the size of each product category in each country. How a sample of shops and homes can measure a whole country is explained in consumer panels and retail audits.
The second group sells what the market knows: information about companies, industries, science and trade. It includes financial research platforms, which search company filings and transcripts of results calls; global trade data, built from shipment records; alternative data, such as card spending that shows how a company's sales are going; and web scraping services, which copy data from public websites. Automotive market data serves the car industry, and citation and patent databases serve science and inventors.
Three more kinds of data stand on their own. Sports data records every pass, shot and play of thousands of matches, for broadcasters, bookmakers and teams. An expert network arranges paid calls with former executives, and sells the transcripts. Software review platforms collect buyers' reviews of business software, and sell them to the companies that make it.
Some large data vendors appear elsewhere on this site: the financial data terminals that investors read, the credit bureaus that hold records of loan repayments, and the property risk data that insurers use to set prices.
Selling data takes a lot of engineering. A dataset that covers a whole market has to be collected from many sources, and the sources have to be matched against each other every day. The business is also changing. More and more, the buyer does not search the dataset at all. An AI data analyst has read the whole dataset, and it answers the buyer's question directly. A subscription to data is becoming a subscription to an analyst.