You can point an AI assistant such as Claude at this article, give it your contact file and your vendor API keys, and have it run the whole evaluation. The steps below are written so that works.
The hidden cost of incorrect emails
The differences in incorrect email addresses are equally important.
In our evaluations, vendors returned anywhere from 21 to 79 incorrect email addresses per 500 contacts when enriching from LinkedIn profiles. That's an error rate ranging from approximately 4% to 16% of contacts, a nearly fourfold difference.
Incorrect addresses aren't just missed sales opportunities. They can generate hard bounces or reach unintended recipients who mark your messages as spam.
If you're sending from your primary company domain, the consequences can extend beyond prospecting to customer communications, recruiting, invoices, and other business-critical emails. For smaller companies with newer sending reputations, a company can effectively lose its ability to deliver outbound email until its reputation recovers.
Choosing the wrong vendor can therefore hurt you twice: fewer prospects reached and greater risk to your entire email operation.
The method: test against emails you know are correct
A vendor trial on a fresh prospect list gives you a pile of addresses and no way to tell which are right. Start from the other end. Take people whose email you already know, ask each vendor to find it, and count how often they do.
Step 1: Build your ground truth
Most B2B companies have useful sources:
- Positive email replies: people who've recently responded to your emails.
- Product logins: users who've verified their work email addresses.
- Closed-won opportunities: contacts your sales team has communicated with directly.
- Existing customers: people your customer success team interacts with regularly.
Use addresses verified within the last 30 days, and confirm each person still works at the listed company. Use work emails only.
Start with approximately 500 contacts, or 1,500 if you want robust metrics across the three major regions (AMER, APAC and EMEA).
Step 2: Prepare the input file
One row per person, with these columns: the known email, the LinkedIn profile URL, first name, last name, company name and company domain. Add a region column if you want results by region.
Step 3: Ask every vendor the same question
Give every vendor the same inputs and never the known email. Test the way you'll use the vendor: by LinkedIn URL if your reps prospect from LinkedIn, by name and company if you enrich CRM records. Run both if you can. A vendor can be strong on one and weak on the other.
Save each vendor's answer for each row: the email it returned, nothing, or an error.
Step 4: Score every answer
Lowercase and trim both emails before comparing. Then each answer is one of four things:
- Correct: the vendor returned exactly the known email.
- Wrong: the vendor returned a different email.
- No email: the vendor returned nothing.
- Error: the call failed. Leave errors out of the rates and report the count.
From those, three measures per vendor:
- Coverage: the percentage of contacts for which the vendor returns an email address.
- Precision: the percentage of returned addresses that are correct.
- Correct-email yield: the percentage of all contacts for which the vendor returns the correct address.
Correct-email yield is the most useful headline number because it captures both coverage and accuracy. Precision matters too, because incorrect addresses damage your sender reputation.
What the output should look like
Vendors ranked by correct-email yield, for each way of looking people up and for each region. If your file has no region column, the Sumble API can identify each contact's region from their LinkedIn profile.
The vendors and numbers below are made up. Switch the lookup method or the region, and click Correct or Wrong to sort.
Correct emailWrong emailNo email returned
Look at the Wrong column as well as the yield. Vendor C finds fewer correct emails than Vendor B and returns twice as many wrong ones.