What’s better: Clearbit, Apollo, Diffbot, or Proxycurl?
I compared these sales data APIs on how well they estimate the number of employees at companies. The inputs for these tools are the company name and domain. The number of employees on LinkedIn served as the “source of truth.” Who performed the best?
The Google Sheets link with the full results is here.
Proxycurl was the clear winner, by a mile. And that’s not surprising; they get their data by scraping LinkedIn directly.
The second place goes to Apollo. It was closest to the actual number 27 times out of 100, and it exactly matched the real number of employees 5 times. It was correct within 10% a total of 30 times. With all the other features that come with Apollo, it’s almost a tie between Proxycurl and Apollo!
Diffbot excels in advanced, custom segmentation and has surpassed Clearbit in estimating the number of employees, securing 3rd place. It’s also the maker of Leadgraph.
Lastly, and disappointingly, Clearbit finished 4th. As a stand-alone tool, it was probably not worth it, but with the Hubspot acquisition, it might still be relevant. What’s your experience with Clearbit? Is it accurate?
Now, in terms of which tool to use… that depends on your acceptable level of error. This is hard to measure; the median metric provided is lacking context. If you need to be super precise, Proxycurl is the answer. Apollo is very solid. I found that when Diffbot is off, it can be off by a lot.
You also have to consider the other parameters you construct the ICP on. For a basic ICP—industry, number of employees, location—you want the employee data to be precise. But when you’re stacking these parameters involving technographic data of tools that only certain company sizes would have, precision on the number of employees is less crucial.