To test a B2B data provider, run the same 200 of your own target contacts through every vendor in the same week, then score three numbers for each: found, verified and correct. A bounce test and a phone connect test give you “correct”, and dividing what you spent by the contacts that passed gives the only price that matters.
This is the bake-off method we use. It fits in about two weeks, needs no special software, and turns vendor claims into numbers from your own market.
How do you test a B2B data provider?
You test a B2B data provider by giving it the same list you give its competitors and checking what comes back with tools the vendor does not control. The list is your real target buyers, the inputs are identical (first name, last name, company domain), and every result is checked by an independent verifier, a real email send and a real call.
The output is one scorecard per vendor: found rate, verified rate, correct rate and cost per usable contact, split by region and by field. Pick on the number that matches how your team sells: correct emails for an email team, correct mobile numbers for a calling team.
Found, verified, correct: what a provider test measures
A provider test measures three layers of B2B data quality, and each layer removes contacts the one before it counted.
- Found: the vendor returned an email or a phone number. Easy to measure, easy to inflate.
- Verified: the email passed a mailbox check, or the number is a live mobile rather than an office switchboard. Risky (catch-all) emails sit in between.
- Correct: the email was delivered and the number reached the named person. Only a real send or a real call proves it.
If you came here to learn how to measure data quality in general, the six classic dimensions still apply. This is how each one shows up in a provider test.
| Dimension | What it means for a contact | How the bake-off measures it |
|---|---|---|
| Accuracy | The email and phone belong to this person | Bounce test, connect test, known-answer set |
| Completeness | The fields you need are filled | Found rate per field (email, mobile) |
| Timeliness | The person still holds that job | Known-answer set, wrong-person calls |
| Validity | Formatted and deliverable | Verification status, phone line type |
| Uniqueness | One record per person | Duplicate rows in each export |
| Consistency | Same values across your tools | Not a vendor metric: fix it in your CRM |
How to test a B2B data provider in 7 steps
The whole test is seven steps, and only steps 5 and 6 need a few days of waiting.
1. Build a 200-contact test list from your real buyers
Pull 200 people who match the accounts you will actually sell to: the titles, company sizes and regions in your plan. If you sell in two regions, build 200 per region and score each one on its own. A 100-contact list is a quick first read, not a verdict.
Do not accept a vendor’s “free data sample”. Cognism, for example, offers one on its site, and any sample the vendor picks will flatter the vendor.
Why 200: a found rate measured on 100 contacts has a margin of error of about plus or minus 10 points. On 200 it is about 7. If vendor A finds 78% and vendor B finds 71% on 100 contacts, you have not proven a difference.
2. Hide a known-answer set of 30 contacts in the list
Add 30 people whose current email and mobile you know are right: recent customers, partners, people you spoke with this month. Mix them in without a label.
A vendor that returns an old employer or a different email for these people is wrong, and you learn it without sending anything. Thirty is a sanity check, not a precise rate, and you never email or call these people as part of the test.
3. Run the same list through every vendor in the same week
Give every vendor the same inputs and nothing extra. Run them within a few days of each other, because people change jobs while you test.
Record one row per contact per vendor. This header works in any spreadsheet:
Log the credits charged on rows where nothing came back. Lusha’s API, for example, uses a minimum of 1 credit per request even when no results are returned, and Apollo’s API docs note that some vendors connected to its enrichment consume credits per lookup even when no data is found.
4. Score found and verified rates with one independent verifier
Found rate is simple: rows with an email returned, divided by rows sent. Do the same for phone numbers.
Then run every returned email through one verifier none of the vendors control, so the labels compare. MillionVerifier, for example, lists 10,000 credits for 39 USD, so checking a few hundred emails costs very little.
- Count Valid, Risky (catch-all) and failed addresses separately. Never count Risky as verified; see our guide to catch-all emails.
- For phones, record the line type. A company switchboard is not a mobile, and the difference decides your connect rate (direct dial vs mobile).
5. Run the bounce test on verified emails
Send one plain, relevant email to every verified address from a warmed secondary domain, never your main one. Count hard bounces within 72 hours; soft bounces are temporary and do not count.
Bounce rate per vendor = hard bounces divided by emails sent. For reference, Hunter reports an average cold email bounce rate of 3.6% across 31 million emails its users sent in 2025.
Send the Risky addresses in a separate, smaller batch so they do not blur the Valid number. Treat the test as real outreach: a genuine reason to write, a working opt-out, and no follow-up to anyone who asks to be removed. For contacts in the EU, read whether B2B data is GDPR compliant first.
6. Run the phone connect test on found mobiles
Call a random 50 of each vendor’s numbers, or all of them if there are fewer. A person dials: calls to a cell number made with an autodialer or a prerecorded message need prior express consent under the Telephone Consumer Protection Act (TCPA).
Call between 8 a.m. and 9 p.m. in the contact’s time zone. The B2B cold calling laws page covers the rest (not legal advice).
Give every call one outcome code:
Correct rate = right person divided by (right person + wrong person + dead number). Leave unanswered calls out of the math and try them twice more. Expect a lot of them: across 894 dials in my own sessions (July to September 2026, mostly Australian tech prospects on direct mobiles), 23% connected.
7. Calculate cost per usable contact and decide
Spend is every credit charged during the test, misses included, times the price per credit, plus any seats you paid for. Usable contacts are the ones that passed: delivered emails for an email team, right-person mobiles for a calling team.
- Cost per usable contact = spend / usable contacts.
- Work it out per region and per field, then check the terms that change the real price: seats, contract length, credit rollover. Our guides to B2B data pricing models and what B2B data really costs walk through both.
Example: a bake-off scorecard for two vendors
Jordan Ellis, RevOps lead at Brightline Freight Software in Denver, compares two vendors for a 6-rep calling team on a 200-contact US list. The numbers below are an example to show the math, not a benchmark.
Vendor A charges 0.50 USD per mobile found and nothing on a miss. Vendor B charges 0.30 USD per lookup, whether or not it finds a number.
| Step | Vendor A | Vendor B |
|---|---|---|
| Mobiles found | 160 of 200 (80%) | 120 of 200 (60%) |
| Spend | 160 x 0.50 = 80.00 USD | 200 x 0.30 = 60.00 USD |
| Right person in the connect test | 85% | 75% |
| Usable mobiles | 160 x 0.85 = 136 | 120 x 0.75 = 90 |
| Cost per usable mobile | 80.00 / 136 = 0.59 USD | 60.00 / 90 = 0.67 USD |
Vendor B’s sticker price is 40% lower, yet each usable mobile costs more, and it leaves 110 of the 200 buyers unreachable against 64 for Vendor A.
What our own benchmark measured
We ran steps 3 and 4 (found only) across 8 providers and 4 regions and published the raw data in our B2B data provider comparison. On the North America list, Scalelist found emails for 92% and phone numbers for 87%; Apollo found phone numbers for 50% there and for 8% in APAC.
It is our own test, and it measures found, not correct. The 2026 B2B data accuracy benchmark explains the limits. Run the rest of this protocol on your own list before you trust anyone’s numbers, ours included.
Common mistakes when you test data providers
Most failed bake-offs are lost in the setup, before a single record comes back.
Testing on the vendor’s sample
A sample the vendor selects shows its best segment. Fix: always send your own list.
Comparing different lists or different weeks
Two vendors tested on different contacts, or a month apart, are not comparable. Fix: one frozen list, one week.
Counting Risky emails as verified
A catch-all domain accepts every address, so the check cannot confirm the mailbox. Fix: report Risky as its own column and send it separately.
Trusting each vendor’s own labels
“Verified” means something different at every vendor. Fix: one independent verifier for all results.
Ignoring what a miss costs
If a vendor charges per lookup, its price per found contact rises as its hit rate falls. Fix: record credits charged on every row, not just the hits.
Testing one region and buying for four
Coverage changes sharply by region. Fix: 200 contacts per region you sell into, scored separately.
Bake-off checklist
Run through this list before you sign anything.
What to look for in a B2B data provider after the test
The test tells you who finds correct data in your market. These terms decide what it costs to keep using it, and they are the criteria behind our list of the best B2B data providers.
- Correct rate in your regions, not a global accuracy claim. Vendor claims such as ZoomInfo’s published “96% Email Deliverability Rate” are measured on their terms, not on your list.
- Charge on success, in writing: the pricing page or terms should say a miss costs nothing.
- Seats: does every rep need a paid seat to see the data?
- Contract: monthly option, or annual only?
- Credit rules: do unused credits roll over or expire?
- Status on every email: Valid versus Risky, visible before you send.
- Line type on every phone: mobile versus office line.
- Compliance documents: a Data Processing Agreement and a published sub-processor list.
How Scalelist fits into your test
Scalelist, an AI lead finder, is easy to put in a bake-off because a miss costs nothing: 1 credit per email found, 20 per mobile found, 0 when nothing is found. Every email comes back labelled Valid or Risky, and Not found is free.
- Test cost: an email-only test on 200 contacts uses at most 200 credits, which fits the 500-credit plan at 29 USD a month. Emails and mobiles for all 200 use at most 4,200 credits (200 x 1 + 200 x 20); at our North America benchmark rates it is about 3,664, which fits the 5,000-credit plan at 199 USD on our pricing page. No seats and no annual contract.
- Before you pay: new accounts get 20 free credits, no credit card required. That is enough for 20 emails or 1 mobile.
- Check other vendors too: the Email Verifier is included and works on addresses from any source.
Run your 200-contact test and see found, Valid and Risky on every row.
Get 20 credits to test itFAQ
How many contacts do you need to test a B2B data provider?
About 200 per region. On 200 contacts a found or correct rate has a margin of error of roughly 7 percentage points; on 100 it is about 10, and on 50 about 14. If you compare two vendors on 100 contacts, a gap under about 14 points can be chance. Add a hidden set of 30 known-answer contacts to check correctness without sending anything.
How long does a data provider bake-off take?
About two weeks. Building the list and the known-answer set takes a day or two, running every vendor takes a few days in the same week, the bounce test needs 72 hours after the send, and the connect test needs a few calling blocks with up to three attempts per number. Scoring and cost math take an afternoon.
What is a good bounce rate when you test B2B data?
As low as your verified set allows, and far below your unverified set. For reference, Hunter reports an average cold email bounce rate of 3.6% across 31 million emails sent in 2025. Measure hard bounces only, within 72 hours, on emails an independent verifier marked Valid, and report Risky catch-all addresses separately.
What is the difference between found and verified data?
Found means the vendor returned an email or phone number. Verified means an email passed a mailbox check or a phone number is a live mobile. Correct means the email was delivered and the number reached the named person. A vendor’s found rate is always the highest of the three, so never compare one vendor’s found rate with another’s verified rate.
Should you trust a vendor’s own accuracy claims?
Treat them as a starting point, not as evidence. Each vendor measures on its own data, with its own definition of accuracy or deliverability, usually on its strongest segment. The only numbers that predict your results come from your list, your regions and an independent check. That is why you test a B2B data provider before you sign, even one with a high published score.
What does it cost to test a data provider?
One month of a plan big enough for 200 contacts, per vendor. Email-only tests fit most entry plans, which cost 29 to 99 USD a month at list prices. Testing mobiles needs more credits: at Scalelist, emails and mobiles for 200 contacts fit the 5,000-credit plan at 199 USD. Add about 39 USD for 10,000 independent verification credits at MillionVerifier, and avoid annual contracts until the test is done.
How often should you re-test your data provider?
Every six to twelve months, and before you expand into a new region or a new buyer type. Contact data changes as people move jobs, so a vendor that was strongest last year may not be now. Our guide to B2B data decay explains how fast lists go stale, and removing duplicate contacts keeps your CRM clean between tests.
