Questions about DataBroker

Yes. There is no account, no email gate and no watermark. The form and the PDF are both made in your browser.

No. Nothing you enter is sent to our servers. Close the tab and it is gone, so download or copy the result before you leave.

Start from your cost to deliver plus a margin, which is the floor. Then look for any quote for similar data, even rough, and use it as your base rate. Adjust for freshness and exclusivity.

Per record suits one-off snapshots. A yearly fee suits feeds that update. You can use the calculator for a yearly fee by pricing one year's licence.

No. The file is read by your browser and scanned there. Nothing is sent to us or stored, and closing the tab clears it.

No. It means none of these patterns matched. Names, addresses and free-text notes can still identify people, so read the file too.

There is no single price. It depends on who is buying, what they can do with it, how fresh it is and whether you have the right to license it. A range from several methods is more honest than one number.

The market method, when you have a real comparable quote. Without one, the cost method gives a floor and the income method gives a ceiling for a specific buyer.

Usually only in part. This one says that only the exclusivity, confidentiality, costs, expiry and governing law clauses bind. The rest records current intent. Courts look at the words used, so keep that clause.

Very little in practice. A term sheet is often a table of terms; a letter of intent is the same content written as a letter and signed. Both lead to a definitive agreement.

For video, photos, audio and books, the estimator multiplies what you hold by prices AI buyers were reported to pay. For other types no price is public, so the honest answer is that it is set deal by deal.

From news reports of specific deals, mostly via Bloomberg, each linked under the result and in Sources. We add a price only when a named outlet published it.

No code is downloaded. Your browser asks GitHub for the repository's public details and language byte counts only, and nothing passes through our servers.

It is an assumption for a fully loaded cost, meaning salary plus benefits, equipment and management. The US median wage for software developers is lower; change the figure to match your market.

Search its name or website in the lookup above. If it is registered in California or Oregon you will see the registration as the state published it. Search Texas's and Vermont's registries on their own sites too.

Yes. California, Texas, Oregon and Vermont all publish their registries. California and Oregon also offer downloadable files, which is what this lookup is built from.

Yes, if it owns the data and may license it. Size matters less than scarcity: a small firm with years of specialist records in a narrow field can hold something a large buyer can't easily collect.

Not if the deal is a licence, which is the usual form: you keep the data and grant the buyer a right to use a copy. An exclusive licence can stop you licensing it to anyone else in the agreed field for the agreed period, so read that clause carefully.

A retailer using sales data to cut waste (improving), a payroll provider showing customers salary benchmarks (wrapping), a logistics firm selling a shipping index (an insight product), and a closed software company licensing its codebase to an AI lab (a data licence).

Not necessarily. MIT's research suggests most companies get more from improving their own work first. Outside sale suits data that is scarce and clean to license, and it carries more legal and reputational risk than internal use.

Not for brokering business datasets under the registry laws we checked. California, Vermont, Texas and Oregon require registration from businesses that sell personal information about consumers they don't have a direct relationship with. If personal data is in your deals, ask a privacy lawyer which rules apply, including any sector-specific ones.

Usually through a commission or finder's fee: a share of the deal value, or a fixed fee, agreed with the seller in writing before you introduce buyers. The rate is negotiated deal by deal, so put it in a signed agreement rather than relying on an email.

Usually not. Code written for an employer generally belongs to the employer under employment terms. Check your contract before offering it to anyone.

Often, if the company still owns it and the people with authority approve the sale. In insolvency, the administrator or court decides. Check investor and lender rights, and remove customer data first.

Charge per record when buyers want different amounts or slices and the unit is easy to count. Charge a flat fee when the buyer takes the whole set and the value lies in having all of it, such as a complete codebase or a full archive.

There is no standard multiple. Price exclusivity against the sales the seller gives up: how many other buyers exist, and for how long the grant runs. Narrow it by time, territory or field of use to keep the premium reasonable.

There is no price list. What a dataset sells for depends on how useful it is to a particular buyer, the rights that come with it and how hard it is to get elsewhere. That is why this guide builds a range from the three approaches instead of quoting a figure per row.

In practice, cost-based methods: adding up what the data took or would take to produce. They are easy to defend but tend to understate value, so sellers usually check them against market comparables and the buyer's likely gain.

No. The right one is where your buyers already work. A buyer on AWS will find AWS Data Exchange easiest; a Snowflake shop will prefer a Snowflake listing. For data that needs explaining, no marketplace replaces a direct conversation.

AWS says providers must be permanent residents or citizens of an eligible jurisdiction, or a business organised there, be registered AWS Marketplace sellers and be qualified by the AWS Data Exchange team. In practice it is built for businesses with data products and customer support.

Mainly investment firms, AI labs and model builders, marketing and sales teams, and product or analytics teams filling gaps. Governments and researchers buy too, under their own procurement rules.

Look for job titles such as data sourcing, data partnerships, data acquisition or alternative data on professional networks, and for people who speak at data industry events. A warm introduction from a broker or a shared contact helps most.

CalPrivacy publishes it on its website. It lists each registered broker with contact details and the disclosures the broker filed, and it is the list DROP uses to route deletion requests.

January 31. Brokers register and pay between January 1 and January 31 of each year for the year after they began brokering Californians' data.

In data work it stands for data as a service. In IT infrastructure the same letters often mean desktop as a service, which is unrelated hosted virtual desktops.

SaaS gives you software to use. DaaS gives you data to use in your own software. Many products mix both, such as a sales database with a search app on top.

Yes, data brokers operate legally in the United States, but they are increasingly regulated. Depending on what they sell, they may need to register in states such as California, Vermont, Texas and Oregon, honour opt-out and deletion requests, and follow the Fair Credit Reporting Act when their data is used for eligibility decisions.

In practice the words are often swapped. Data provider usually describes a company selling its own compiled product, such as company contact data. Data broker is broader and includes intermediaries who resell or arrange the sale of data owned by others.

Labelled photos for object recognition, transcribed audio for speech models, pairs of questions and answers for chatbots, historical transactions marked as fraud or not, and code repositories with commit history are all common types.

It depends on the task and on whether the buyer is training from scratch or adapting an existing model. Adapting a model to a narrow task can work with far fewer examples than building one, which is why small specialist collections still find buyers.

For some jobs, yes: labs report generating most of their alignment examples and large amounts of reasoning practice. For factual knowledge and rare cases, research and lab reports say models trained only on synthetic data do worse, so real data is still needed.

The loss of quality that happens when models are trained over and over on content produced by earlier models. A 2024 Nature paper found that rare cases from the original data disappear first. Keeping real data in the mix appears to prevent it.

Adding an industry category and company size to a list of business accounts, attaching local weather to daily store sales, or tagging support tickets by topic are all forms of enrichment.

Cleansing corrects or removes what is already in a dataset: typos, duplicates, invalid values. Enrichment adds something new. Most projects clean first, because matching dirty records to outside sources gives poor results.

No. Plenty of it describes machines, stock, routes or prices rather than people. Customer records are personal, but a factory's yield logs or a carrier's depot timings may contain nothing about any individual.

Often, but not automatically. It depends on what customers were told, what client contracts say and which privacy laws apply. Non-personal operational records are usually the easiest to license.

It depends on your market. Coverage differs by country, industry and company size, so the only reliable answer is a blind test on a sample of accounts you already know.

Generally yes in the United States, with conditions on how you use it: CAN-SPAM for email, TCPA and do-not-call rules for phones, and state privacy rights for the people named. In the EU and UK you also need a lawful basis under GDPR.

From scraped public web content, licensed archives from publishers and platforms, open datasets, synthetic data their own models generate, and paid human work such as writing, labelling and rating answers.

Often, if it owns the rights and the data isn't restricted by customer contracts, privacy notices or third-party licences. Those three checks decide most deals.

It is a probability, not a fact. Mapping anonymous reading to a company is imperfect, especially for remote workers and small firms. Test it against your own closed deals before relying on it.

First-party intent comes from your own website, ads and emails. Third-party intent is licensed from someone else and shows research happening on sites you don't control.

Firmographic data describes the company itself: industry, size, revenue and location. Technographic data describes the technology it uses, such as its CRM, cloud provider or analytics tools.

Aggregate counts, yes: Census County Business Patterns is free. Named company records with size and revenue estimates for private firms are mostly paid.

It attracts people interested in a product, collects their details or connects their call, and sells that contact to businesses that want customers, per lead, per click, per call or as a commission on a closed sale.

Most large ones don't publish prices. They sell through real-time auctions, account managers or memberships, and the price depends on the vertical, the targeting and whether the lead is exclusive.

Published figures range from about $20 to $60 per contact in many markets to several hundred dollars in competitive metros, and Zillow's reported averages in 2026 were $223 in major metros and $139 elsewhere. Pay-at-closing programmes instead take roughly 15% to 40% of the commission.

Selling contact data at a flat price is treated differently from taking a share of a commission, which most states restrict to licensed people. Check your state real estate commission's rules or ask a lawyer before agreeing any fee tied to a closing.

Published ranges run from about $5 to $45 for shared web leads, $75 to $150 for exclusive leads, and $50 to over $200 for live transfers, depending on the source and the targeting.

Sometimes. Exclusive leads usually reach the consumer more often, but whether they win depends on the agent's close rate and average premium. Compare cost per issued policy, not price per lead.

Lists built from UCC-1 financing statements, the public filings lenders make when they take a security interest in a business's assets. They show a business has borrowed before, but they carry no consent to be called.

Merchant cash advance funders and ISOs, the independent brokers who gather a merchant's documents and shop the deal to funders for a commission.

ActiveProspect's published benchmarks run from under $10 for aged records to $25 to $100 for shared leads, $100 to $250 for exclusive leads and $150 to $500 for booked appointments. Prices move with location, homeownership checks and exclusivity.

Per unit, yes, because more of the selling is done. Whether they are cheaper per install depends on the show rate, which is why installers ask for held-appointment figures rather than booked ones.

Published 2026 ranges run from $15 to $75 for shared marketplace leads, $18 to $80 per valid Local Services Ads call and $40 to $150 or more for exclusive pay-per-lead. Replacement enquiries cost more than repair calls.

For repairs, usually yes, because the homeowner wants someone now and books the first company that answers. Replacement projects tolerate a slower follow-up, so form leads and booked estimates work there.