Buying leads means paying a third party for contact records or for introductions to people who match your customer profile, instead of generating them through your own marketing. In 2026 there are five places to do it: subscription databases, list brokers, file marketplaces, pay-per-lead networks and industry directories. They differ most in freshness, and freshness decides the bounce rate: across 6.5 million non-freemail (business and ISP) addresses BounceZero verified this year, one in seven was already dead.
Cost figures are vendor-reported 2026 ranges. The bounce column is what independent verification of exports typically finds, from the BounceZero corpus and customer uploads, before any cleaning.
| Where | How it works | Cost (vendor-reported) | Freshness | Invalid before verification |
|---|---|---|---|---|
| Subscription database | Search and export contacts by filter; monthly or annual seat | $50 to $1,500 per seat per month | Re-crawled continuously; rows still age | 10 to 22% |
| List broker | You describe the segment; they compile and deliver a file | $0.15 to $0.60 per contact | Depends on broker; ask per row | 12 to 30% |
| File marketplace | Download a ready CSV by category | $0.01 to $0.10 per contact | Often years old | 25 to 45% |
| Pay-per-lead network | You pay for a form fill or call that matches criteria | $20 to $300 per lead | Minutes old | Low, but 5 to 10% typos and throwaways |
| Industry directory | Trade body, chamber, registry listings | Free to low | Varies; registries are good | 15 to 35% |
Pay-per-lead networks sell intent, not files. The other four sell data. Different problems, different checks.
A subscription database suits a team that prospects every week and wants filters, not a file. The seat cost is the entry fee; the per-contact cost falls with use. ZoomInfo, Apollo, Cognism and Lusha dominate this category, and their exports are compared on accuracy in the provider roundup.
A broker suits a one-off campaign to a segment a database does not index well, such as independent pharmacies in two counties or plant managers at food processors. Brokers vary enormously; the checklist below is written with them in mind.
A file marketplace is where most bad purchases happen. The price is attractive because the data has been sold many times already; by the time it reaches you, a third of it may be gone. Buy here only with a sample test and only for throwaway experiments.
Pay-per-lead networks are a different product. You buy a person who raised a hand, usually on a comparison site. The data is fresh but the lead is shared with competitors, and the form-filled address still needs validation: 5 to 10% of inbound addresses have typos or are throwaways.
Directories are the underused option. Registries such as Companies House and professional bodies publish contact data that is accurate for the organisation even when the named person has moved on. Pair them with a finder and a verifier and you have a list nobody else is mailing.
Whichever vendor you choose, the country of the target decides much of the bounce rate. These are BounceZero corpus results for business domains by country ending, April to October 2026.
| Country ending | Verified | Invalid | Catch-all | Unknown | Confirmed valid |
|---|---|---|---|---|---|
| .uk | 21,194 | 17.0% | 3.7% | 10.6% | 65.0% |
| .ca | 7,561 | 15.9% | 6.2% | 13.9% | 54.4% |
| .nl | 87,309 | 21.9% | 0.2% | 4.5% | 54.6% |
| .fr | 119,582 | 12.1% | 0.5% | 9.3% | 47.6% |
| .au | 560,378 | 36.5% | 6.3% | 17.6% | 37.3% |
| .us | 10,998 | 34.1% | 13.6% | 21.9% | 29.1% |
| .de | 2,095,440 | 13.9% | 0.0% | 41.2% | 16.5% |
Australian files are the worst buy in the corpus; the ISP mailboxes that dominate them (hotmail.com.au 92.7% invalid, tpg.com.au 30.1%) have been abandoned in bulk. German unknowns are a probe-refusal issue at t-online.de, not dead mailboxes. Full tables: Lead List Quality Benchmarks 2026.
Eight questions, in the order that eliminates vendors fastest.
If the answer is no, stop. If the sample is "top rows" or hand-picked, ask again for random. Verify it before anything else; the thresholds are in the next table.
A date per row and a method (SMTP check, observed send, pattern). Rows older than 90 days should be discounted or re-verified by the vendor.
Inferred rows are pattern guesses. They bounce two to three times more often than observed rows and are invisible in a CSV unless flagged.
Both should be flagged or excluded. Catch-all rows cannot be confirmed; role rows (info@, sales@) rarely reach a buyer and draw complaints.
Thirty days and a bounce log is standard among reputable brokers. Get the threshold and the process in the contract.
For UK and EU contacts, the vendor must show legitimate interest or consent and a privacy notice naming them as source. For US contacts, CAN-SPAM compliance on your side suffices.
Paying for people you already have, or for your own staff, is common. Send a hashed suppression file.
Single-use is cheaper and often enough. Multi-use means you can re-verify and re-send without re-buying, which matters for sequences longer than one campaign.
Run the sample through verification and compare against these cut-offs. They are deliberately tighter than corpus averages, because you are choosing which file to buy, not rescuing one you already own.
| Measure | Buy | Negotiate | Decline |
|---|---|---|---|
| Invalid share | Under 5% | 5 to 10% | Over 10% |
| Catch-all plus unknown (non-German) | Under 8% | 8 to 15% | Over 15% |
| Role addresses | Under 3% | 3 to 8% | Over 8% |
| Freemail in a business file | Under 5% | 5 to 15% | Over 15% |
| Duplicates | Under 1% | 1 to 3% | Over 3% |
| Disposable or trap flags | None | One in 200 | More |
A 200-row sample fits inside the 100 free checks a month on two accounts, or the $6 Starter pack. Unknowns are refunded.
Because these vendors sell a person who filled a form, the data checks are replaced by terms checks.
Shared leads go to three to five buyers. Exclusive leads cost two to four times more and convert far better. Know which you are buying.
Wrong numbers, duplicates and out-of-area leads should be returnable within a few days. Networks that refuse returns are pricing in junk.
Shared leads are won by whoever calls first. If you cannot respond within five minutes during business hours, buy exclusive or do not buy.
Even hand-typed addresses fail 5 to 10% of the time. Validate the lead's email through the API at the moment it arrives so follow-up mail is not the first bounce.
Which site, which form, which question. A lead that answered "just browsing" is priced like one that asked for a quote unless you check.
Daily caps protect you from a runaway campaign on the network's side. Set them low until return rates are known.
For any segment you can describe by title, company size, industry and geography, the finder-plus-verifier route costs about the same per contact as a broker and produces fresher data. A finder returns a probable address for a named person at a company; a verifier confirms it now. The email finder guide compares the tools, and the list building guide walks the process.
The practical split most teams land on: buy from a database for the broad middle of the market, build by hand for the fifty accounts that matter most, and never touch a marketplace file without a sample test.
The 12-point check and what bought files bounce at
Subscription databases compared
Corpus results by provider and country
Lead types, sources and qualification
Lawful basis, notices and opt-outs for bought data
Every source ranked by data quality
For ongoing prospecting, a subscription database such as Apollo, Cognism or ZoomInfo, because filters and refresh beat a static file. For a one-off niche segment, a broker who will provide a random sample and a bounce replacement clause. File marketplaces are the cheapest and the worst; buy there only for experiments and only after a sample test.
Vendor-reported 2026 ranges: a cent to ten cents per contact on marketplaces, fifteen to sixty cents from brokers, fifty to a thousand dollars a month per seat for databases, and twenty to three hundred dollars per lead on pay-per-lead networks. The cheaper the data, the more of it is dead.
Get a random sample of 200 rows and verify it. Under 5% invalid, under 8% catch-all plus unknown, under 3% role addresses and no spam-trap flags is a good file. Over 10% invalid means the full file will bounce past the 2% threshold Gmail and Microsoft enforce.
When the segment is clear, the file is fresh and you verify before sending, yes: the cost per confirmed contact is comparable to building the list yourself and you have it this week. When the file is a recycled marketplace CSV, no: a third of it is gone and the rest has been mailed by everyone who bought it before you.
Yes for corporate contacts under PECR, provided you identify yourself, offer an opt-out and have a documented legitimate interest under UK GDPR, with a privacy notice that names the data source. Sole traders and partnerships are treated as individuals and need consent. Ask the vendor to show the lawful basis for each record.
Random 200 rows in, valid, invalid, catch-all and unknown out. Decide with numbers, not the vendor's deck. 100 free checks a month.
Ayoub built BounceZero's 5-stage validation pipeline, its dedicated BGP-announced IP infrastructure, and the Patroni HA PostgreSQL cluster behind every verification. Previously built high-volume email delivery infrastructure. Trained at 1337 Benguerir (École 42 network, 2019). Open-source: bgp_analyzer.
Where lists come from, what bounces, and what to check before you pay
How databases are built and where they decay
Thinking of buying a B2B email marketing list? Corpus data from 7.93M verified addresses shows what...
Buy leads for real estate in 2026
Evaluating a B2B database for sale? Compare subscription databases and one-off files on coverage, accuracy...
Business leads lists for sale in 2026
Real estate email address lists in 2026
Continue through related topics