Buying leads as an ISO

01Start from your floor, not the catalogue

Count your callable hours before you look at a price list. Agents multiplied by dials per hour multiplied by hours worked gives you a ceiling, and buying past that ceiling is not ambition, it is waste.

Six hundred, not ten thousand. A file bought at ten times your capacity does not sit patiently — it ages while it waits, and by the time you reach the bottom of it you are dialing data you paid fresh prices for. Buy to the ceiling and reorder.

02Choose record depth by what your team actually does

• A floor that only dials needs phone contacts. Verified email addresses on those records are a line item you are paying for and never opening.
• A team running email or SMS sequences needs the email file, and needs it bounce-checked before delivery rather than after.
• A shop that pre-screens before dialing needs full records — revenue, time in business, amount requested — because the underwriting filter is what makes the smaller call volume pay.

Paying for columns nobody opens is the quietest form of overspending in this business, because it never shows up as a failure. It just makes every funded deal cost more than it should.

03Test at the smallest readable volume

Readable means enough records to expect around ten decision-maker conversations. At a 20% contact rate that is roughly fifty records. Most vendor minimums sit comfortably above that, which is fine — the point is not to buy as little as possible, it is to resist buying ten thousand to see how it goes.

Ten conversations will not give you a reliable funded-deal rate. They will tell you whether the phone numbers connect, whether the merchants match the description, and whether your script survives contact. Those three answers are worth more at the start than a precise conversion figure you cannot yet trust.

04The four numbers

Per cohort, per source, per week:

1. Dial attempts. The denominator for everything else.
2. Decision-maker conversations. Not connections — conversations with someone who can say yes.
3. Applications submitted. The first number that reflects lead quality rather than dialing effort.
4. Deals funded. The only one that pays for anything.

Cost per submission moves within days and is what you use to steer. Cost per funded deal takes weeks to stabilise and is what you use to decide. Teams that track only the second one make decisions a month late; teams that track only the first scale sources that never fund.

05Scale on evidence, not on a good week

Increase volume on a source only after it has cleared your cost-per-submission threshold in two consecutive cohorts. One good week is variance, and variance on small samples in this business is large enough to be genuinely misleading.

The corollary matters just as much: one bad week is also variance. Floors that cut a source after a single disappointing cohort end up cycling through vendors permanently, never accumulating enough data on any of them to know what actually works.

06Dedupe, without exception

Keep a master list of every record your floor has ever dialed. Send it to the vendor before each order, or scrub on receipt.

Buying the same merchant twice costs you three times: the data, the agent time spent redialing someone who already declined, and the merchant's patience. It is the most common silent waste in lead buying precisely because nothing in the process flags it — the file looks fine, the dials happen, and the outcome is quietly worse than it should be.

07Vendor red flags

• No opt-in or filing date per row. Without it you cannot audit age, and age is the claim most often stretched.
• Buyer count described rather than numbered. Limited distribution is not a number.
• Vagueness about where phone numbers came from. Especially on UCC data, where the registry contains no numbers at all.
• No replacement policy, or a claim window shorter than a week. Disconnects are normal; refusing to stand behind them is not.
• A monthly contract before you have tested a single file.
• Conversion claims with no stated denominator. A close rate means nothing without knowing what it is a percentage of.
• Compliance claims that are out of date. A vendor still citing one-to-one consent as a live federal requirement is working from 2024 information.

08Common questions