Benchmark
How accurate is email enrichment? The waterfall benchmark
Last updated
We gave 6 data vendors the same 986 emails and checked every answer against what the person told us. The best order to ask them in depends on the email.
- Vendors tested
- 6
- People checked
- 986
- Waterfall orders tested
- 1,956
- Forms filled
- May 2024 to Sep 2026
Enrichment turns an email address into a name, a company and a job title. Most teams pay for more than one data vendor and ask them in turn until one has an answer. That chain is called a waterfall. We wanted to know how often the answer is right, and which vendor to ask first.
Read this first
- This is our data, not yours. Every person here filled in a form on a B2B website, and most of those websites are tech companies. A list of retail shoppers, doctors or factory managers would give different numbers. Treat these results as a guide, and test vendors on your own contacts before you pick one.
- Accuracy matters more than fill rate. A blank tells you to look elsewhere. A wrong company sends the lead to the wrong rep, the wrong account and the wrong email, and nobody notices. So we rank vendors by accuracy, and a high fill rate is only good news when the answers are right.
What we found
Vendors report fill rate: how often they return a value. We add accuracy: how often that value matched what the person told us. Below, a result counts as right only when both the person's name and their company are right.
- Work emails: mostly right. A common setup asks FullEnrich, The Swarm, Apollo, Clearbit and then FullContact, in that order. It filled the name and company for 66% of work emails, and 85% of those were right.
- Personal emails: mostly wrong. For Gmail, Yahoo and other personal emails, the same setup filled the name and company for 49%, but only 17% of those were right. The name alone was right 80% of the time. Vendors find the right person, then name an old or different employer.
- The order matters, not just the vendors. We tried the same 6 vendors in all 720 possible orders. Accuracy ranged from 73% to 79%, depending only on the order. The common setup above named the wrong company more often than 110 of the 119 other orders of the same 5 vendors. The most accurate order for work emails is not the most accurate for personal ones.
- Job titles are rough. On their own, vendors got the title right 50% to 61% of the time for work emails, and 26% to 35% for personal ones, even counting close matches.
Each vendor on its own
How each vendor did when we asked it alone. Fill rate is how often it returned a name and company. Accuracy is how often both were right.
| Work emails | Personal emails | |||
|---|---|---|---|---|
| Vendor | Fill rate | Accuracy | Fill rate | Accuracy |
| The Swarm | 12% | 94% | 13% | 28% |
| MoltSets | 33% | 89% | 26% | 18% |
| Apollo | 60% | 88% | 23% | 18% |
| Clearbit | 23% | 87% | 11% | 29% |
| FullContact | 14% | 84% | 18% | 17% |
| FullEnrich | 50% | 84% | 47% | 17% |
By field
The same vendors, field by field. On personal emails the name is usually right and the company usually is not.
| Company | Name | Title | ||||
|---|---|---|---|---|---|---|
| Vendor | Fill rate | Accuracy | Fill rate | Accuracy | Fill rate | Accuracy |
| The Swarm | 11% | 97% | 12% | 98% | 11% | 61% |
| MoltSets | 33% | 91% | 33% | 96% | 33% | 57% |
| Apollo | 60% | 93% | 67% | 92% | 50% | 58% |
| Clearbit | 23% | 92% | 24% | 95% | 22% | 55% |
| FullContact | 14% | 90% | 56% | 64% | 15% | 50% |
| FullEnrich | 51% | 88% | 51% | 94% | 51% | 57% |
| Company | Name | Title | ||||
|---|---|---|---|---|---|---|
| Vendor | Fill rate | Accuracy | Fill rate | Accuracy | Fill rate | Accuracy |
| Clearbit | 11% | 36% | 28% | 88% | 14% | 33% |
| The Swarm | 13% | 27% | 16% | 94% | 14% | 34% |
| Apollo | 24% | 20% | 63% | 80% | 23% | 35% |
| MoltSets | 27% | 19% | 31% | 90% | 28% | 29% |
| FullContact | 19% | 17% | 51% | 74% | 20% | 26% |
| FullEnrich | 47% | 18% | 55% | 90% | 50% | 27% |
Speed differs too. FullEnrich took a median of 12 seconds per email. Every other vendor answered in under 1 second.
Build your own waterfall
Pick vendors and put them in order to see what that waterfall would have found for these 986 people. Every dot in the chart is one of the 1,956 orders we tested, placed by fill rate and accuracy for the name and company together.
Loading the builder…
What it takes to run a waterfall
Buying the vendors is the easy part. Keeping the waterfall right is the hard part. This test shows why.
- Order changes the answer. The same 6 vendors named the wrong company anywhere from 88 to 135 times, depending only on the order. Change one vendor and every order has to be tested again.
- One order does not fit every email. On their own, vendors were 88% to 97% accurate on the company for work emails, but only 17% to 36% for personal ones. Picking the order by email type raised company accuracy from 90% to 92% on work emails, and from 18% to 29% on personal ones. On personal emails it filled fewer companies (50% down to 23%). Per 100 personal emails, that meant 7 right companies instead of 9, and 16 wrong ones instead of 41. A blank is better than a wrong employer.
- The vendor that fills the most is not the most accurate. On work emails, Apollo had the highest fill rate for the company (60%). The Swarm was the most accurate (97%), with a fill rate of only 11%.
- Cost and speed move with the order. Some vendors charge per lookup, some per match and some a flat fee. Moving a flat-fee vendor earlier cut paid lookups on work emails by 10%. FullEnrich took a median of 12 seconds per email, so where it sits decides how long you wait.
- Fill rate hides mistakes. A vendor with a higher fill rate looks better on a dashboard. Only a list of people whose company you already know shows its accuracy.
- It drifts. People change jobs, vendors update their data and their prices, and accounts run out of credits. While we ran this test, one vendor's account ran out of credits. It stopped finding anyone, and nothing in the waterfall flagged it.
Method and caveats
- Where the right answers came from. Each person typed their name, company and job title into a business web form. A company counts as right when its name or website matches, including close variants (“Acme” and “Acme Robotics”). For a work email, the company that owns the email's domain also counts as right.
- What we cleaned up. We fixed obvious typos and kept the person's own words. We left out values that are not a job, an employer or a name, such as “Resident”, a street address or a single letter. We changed a name only when the email address itself spells a different one.
- Why 986 and not 1,000. We started with 1,000 people. For 14 of them, the name on the form did not match the email address, or the address was a shared inbox. We could not tell whose email it was, so we left them out.
- Each rate counts only people who told us that field. 961 people gave a company we could check, 971 a name and 970 a job title. “Name and company” counts people who gave both, and is right only when both are.
- We checked the matching by hand. Matching is automatic. We read 30 “wrong company” answers ourselves: 28 were wrong, and 2 were the same company written another way (“UCSB” and “UC Santa Barbara”). So accuracy may be a little higher than shown.
- People change jobs. The forms were filled between May 2024 and Sep 2026. We asked the vendors in September and October 2026, so some “wrong” answers may be a newer job. That affects every vendor the same way, and vendors were no more accurate on recent forms.
- LinkedIn follow-ups are included. When a vendor returns a LinkedIn profile, the waterfall looks up that profile to fill in the job title.
- Who is in the sample. These are people who fill in forms on B2B company websites, most of them tech companies. 219 of the 986 used a personal address. Your own contacts will look different.
- No prices. Vendor prices depend on your plan and volume, so we left them out.
- Small differences are noise. A gap of a few answers between two orders could be chance. The gaps we call out held up when we split the sample in half.