# How to Buy B2B Contact Data with a 100-Account Sample

A method for comparing contact data providers on your own accounts: missing, wrong and usable rates by region, cost per usable record, and what the result means when the thing you are replacing is ZoomInfo or Cognism.

By Fama. Published 2026-11-10. Updated 2026-11-11. Editorial analysis; examples are illustrative unless stated otherwise.

Two data vendors have sent you sample files. One has more rows. The other has more phone numbers. If the vendors chose the accounts, both samples may favor the markets they cover best. If you choose from those files you are choosing the better sample-builder.

A better evaluation begins with your own sample: you pick the accounts, you define what counts as usable, and you do the checking. What follows is that method, worked through with invented numbers, and then what it does and does not tell you about replacing ZoomInfo or Cognism specifically.

## Define Usable Before You Look at Any Data

Pick one target role per account, the person you would actually contact first. Then sort every account into exactly one of three outcomes.

**Missing**: the provider returns no one in the target role at that company. **Wrong**: it returns someone, and on checking, the person has left, holds a different role, or no contact route you intend to use works. **Usable**: the person is currently in the role and at least one route you intend to use works.

Keeping missing and wrong apart matters because they cost different things. A missing record costs you an account you cannot reach. A wrong record costs a credit, a seller's time and sometimes a message sent to the wrong person. An email address that passes a verification tool provides a signal about deliverability, with limits depending on the check; it says nothing about whether its owner still runs finance.

## Build the Sample Yourself

Choose 100 accounts from your real target list, split by the regions you sell into in rough proportion to where you want pipeline. One hundred is a planning choice: large enough to show big differences, small enough for one person to verify. It is not a statistically validated size, and the strata below are smaller still.

Include awkward cases on purpose: subsidiaries, recently reorganized companies, small firms. Give each provider identical inputs, a company name and domain, and ask them to process the same full sample without substituting easier accounts.

## A Worked Sample

Suppose a hypothetical team sells to heads of finance and wants growth in Germany. Its sample is 40 US accounts, 30 UK and 30 German. Two providers, called A and B here and standing for no real product, return the following. Each cell reads missing / wrong / usable.

| Stratum | Provider A | Provider B |
| --- | --- | --- |
| US (40) | 6 / 6 / 28 | 10 / 4 / 26 |
| UK (30) | 9 / 6 / 15 | 6 / 3 / 21 |
| Germany (30) | 15 / 6 / 9 | 9 / 3 / 18 |
| All (100) | 30 / 18 / 52 | 25 / 10 / 65 |

Overall, A is usable for 52 of 100 accounts and B for 65. That headline hides the regional pattern. A's usable rate is 70% in the US (28 of 40), 50% in the UK (15 of 30) and 30% in Germany (9 of 30). B's is 65%, 70% and 60%.

Now look at error among the records each provider did return. A returned 70 and 18 were wrong, which is 25.7%. B returned 75 and 10 were wrong, 13.3%. That rate drives your checking workload and your exposure to misdirected outreach.

## Cost per Usable Record

Price the sample as if it were production. Assume, hypothetically, that A's quote works out at $2 per returned record and B's at $3, that both charge for wrong records, and that checking a returned record takes three minutes of someone whose loaded cost is $40 an hour, so $2 of labor each.

A: 70 returned × $4 = $280, divided by 52 usable, is $5.38 per usable record. B: 75 returned × $5 = $375, divided by 65 usable, is $5.77.

A is cheaper per usable record overall. By stratum the answer moves. In the US, A costs 34 × $4 = $136 for 28 usable, or $4.86, against B's 30 × $5 = $150 for 26, or $5.77. In Germany, A costs 15 × $4 = $60 for 9 usable, or $6.67, against B's 21 × $5 = $105 for 18, or $5.83.

Cost per usable record still understates the German difference, because it prices only the accounts a provider can reach. A leaves 21 of 30 German accounts without a usable contact (15 missing plus 6 wrong); B leaves 12. If Germany is where next year's pipeline has to come from, the unreachable accounts are the larger cost and no per-record figure shows it.

The working conclusion for this sample is to investigate B for Germany and the UK. The US difference is only two usable accounts; it is too weak a basis here for claiming a coverage advantage. Price, contracts and a fresh sample deserve weight. Whether to run two providers depends on whether someone will own deduplication between them.

## What the Sample Does Not Prove

Thirty accounts in a stratum is thin. A gap of 9 against 18 warrants follow-up; a gap of two gives little practical reason to switch on its own. Neither difference has been tested for statistical significance. The sample covers one role, one moment and the accounts you happened to choose. It supports a purchasing decision for your list. It does not establish a provider's accuracy in a country, and it should not be quoted as if it did. If a contract is large, rerun a fresh hundred before signing instead of extrapolating from the first.

## If You Are Replacing ZoomInfo

Work out what you would actually be removing. [ZoomInfo](https://www.zoominfo.com/) presents GTM data alongside prospecting, ABM, deal management, enrichment and orchestration, expansion and conversation intelligence. A team that uses only contact lookup can treat replacement as the data test above. A team whose marketing runs account-based programs from it, or whose CRM enrichment and conversation intelligence depend on it, is replacing several systems, and a contact-data vendor that wins the sample will still leave those jobs uncovered. List every workflow that reads from it before you compare anything.

## If You Are Replacing Cognism

[Cognism](https://www.cognism.com/) presents sales intelligence with phone-verified contact data, signals, CRM enrichment and access through an API and MCP. Two things follow for scope. If your sellers rely on phone outreach, report a separate phone-usability measure: of the usable records, how many have a number that reaches the named person, checked the way you are entitled to check it. And if enrichment jobs or internal tools call the API, the replacement has to serve those consumers as well as the people searching by hand. Regional fit, in either direction, is something only your own sample can show.

## When This Method Is the Wrong Tool

If you sell to a few hundred named accounts, a researcher building the contact list by hand may beat any subscription, and the sample will mostly tell you so. If the real problem is that sellers contact the right people with the wrong reason, better data changes nothing. And before signing with anyone, ask in writing what consumes a credit when a lookup fails or returns a duplicate, and what you may keep when the contract ends; both change the arithmetic above.

---

[Editorial methodology](https://usefama.com/methodology)