Illingworth says Apollo-only lists run 50-65% valid
One operator's build order for lists, plus the bounce thresholds he uses to decide when to stop sending.
Richard Illingworth says raw Apollo exports come back only 50-65% valid, and that most cold email bounces trace back to where the list came from rather than anything in the copy or the infrastructure.
His fix is a layered build. Pull the actual ICP from Google Maps via Apify instead of a stale Apollo export, run Findymail to fill in the addresses Apollo left blank, validate every record through Million Verifier, then export the clean file and only then load it into the sequencer. The order matters in his framing because the sequencer is the last step, not the place where filtering happens.
The numbers he checks after the wash
Illingworth puts a healthy clean rate at around 75%. Wash a 10,000-contact list and you keep roughly 7,500 usable records. If bounce creeps over 2%, he stops and re-verifies before sending again, and he wants every list verified within three days of sending because lists go stale fast.
The more useful part is the inverse check. A verifier reporting 90-95% clean is not good news in his read, it is a signal the tool is passing records it should not, and the answer is to re-run the file through a second verifier. That flips the usual instinct, where a clean report is treated as permission to start sending.
Why he rejects Apollo-only lists
In a separate post, Illingworth says he has never seen an Apollo-only list book real meetings. His reasoning runs past deliverability. A big chunk of the emails are stale and bounce, the good contacts already got hit by 50 other people that week, and with no enrichment every email reads like a broadcast. The bounce rate then drags domain reputation down, so the list problem becomes an infrastructure problem.
i've NEVER seen an Apollo-only list book real meetings.
He frames the cost side simply, saying the process costs less than replacing burnt domains.
For people sending, this is one operator's benchmark set and not a dataset. He does not say how many lists he has washed or over what period, so treat 50-65% and 75% as his working numbers rather than an industry figure. The part that transfers regardless of your stack is the second-verifier habit. If you are running a single validation pass and taking the score at face value, a high clean rate is the one result you have no way to falsify, and the bill for getting it wrong shows up as a bounce rate over 2% and domains you have to replace.
if a verifier reports 90-95% clean, that's a red flag. re-run it through a second tool.