Signal Index #1: Where Verified B2B Contacts Actually Come From
Which discovery method produces the most verified B2B contacts?
Google discovery with our own verification produced a verified email for 22.8% of businesses found (95% CI 21.3-24.4%, n=2,839). Paid email-finder APIs produced 18.3% (16.9-19.8%, n=2,895). Community scrapes of Hacker News and Product Hunt produced 5.4% (4.7-6.3%, n=3,260).
That first result is the one worth sitting with. Paying an API per record returned fewer usable contacts per business discovered than searching Google and verifying the addresses ourselves. The intervals do not overlap, so this is not noise.
Where does this data come from?
From our own outbound pipeline: 9,036 businesses discovered between roughly May and August 2026 by SignalEngine, of which 1,304 were actually emailed. Every figure is computed by an open script from a raw database dump, and every percentage carries a 95% confidence interval and the sample size it came from.
This is not a survey and not a vendor benchmark. It is one pipeline's operational record, which is both its strength and its limit: the numbers are real and unedited, and they describe UK-weighted B2B agency and services prospecting rather than the whole market.
Why do paid email-finders underperform on yield?
Because the two methods are measured against different denominators in practice. A paid finder returns an address for essentially every record you submit, but only 18.3% of those addresses cleared verification. Google discovery finds fewer addresses overall, and a much higher share of the ones it does find are real.
Put plainly: buying a contact list is buying rows, not buying reachable people. The gap between "has an email attached" and "has a verified email attached" is where the money goes. Of the 2,895 businesses sourced through paid finders, 100% had an email address attached and 18.3% had one that verified.
What is the worst-performing discovery source?
Community scrapes, by a wide margin: 5.4% verified-contact yield against 22.8% and 18.3% for the other two methods. Hacker News and Product Hunt accounted for 3,260 of 9,036 businesses in the dataset — 36% of everything discovered — and produced the fewest reachable contacts of any method.
This is the finding with the most immediate operational value. Just over a third of the pipeline's discovery volume was going into the source least likely to yield anyone you can email. Volume of discovered businesses is a vanity number if the contact-find rate is not measured alongside it.
What is NOT in this issue, and why?
Reply rates. The dataset contains 15 replies in total, which is far below the 30-observation minimum this publication uses for any reported cell. No reply-rate figure appears at any breakdown, and anyone quoting one from this dataset is inventing it.
The bounce-rate comparison is also withheld. Paid email-finders bounced 6.8% (5.5-8.5%, n=1,100) and Google discovery bounced 1.9% (0.5-6.7%, n=104). That looks like a dramatic difference and it is not yet a finding: the Google arm rests on 2 bounces from 104 sends, and the confidence intervals overlap. It was the most quotable sentence available in this dataset, and it did not survive the test.
How is "verified" defined here?
A contact counts as verified only when our verification cascade confirmed the address, never when a model or a data vendor asserted it. A business counts once regardless of how many contacts it has, so the percentages are "share of businesses that yielded at least one verified email", not "share of email addresses that verified".
Bounce rate is bounced emails over emails actually sent. Prospects still queued or never contacted are excluded from that denominator — 7,732 of the 9,036 had not been emailed at the time of the extract, and counting them as sends would have deflated the bounce rate roughly sevenfold.
How should a solo founder use this?
Measure your own contact-find rate before buying anything. Take 100 businesses from a source, run them through verification, and count how many produce a reachable address. That single number tells you more about a source's value than its price per record, and most people never calculate it.
If you are choosing where to spend, the defensible conclusion from this dataset is narrow: on this pipeline, for this audience, search-plus-verify beat pay-per-record on yield, and both crushed community scraping. Whether that holds for your audience is an empirical question you can answer in an afternoon. See [the best AI cold outbound tools for a solo founder](/learn/best-ai-cold-outbound-tool-solo-founder) for what sits around the data layer.