Two months of lead automation: why only 64 of 809 leads got an email
809 fetched, 64 sent, 10 replies. The filter in the middle is what the system is actually for.
The numbers at each step
gig-radar pulls leads from several sources, de-duplicates them, has a model score them, drafts per channel for the ones above the line, a person reviews the last step before anything is sent, and replies are scanned automatically. Over two months: 809 fetched, 64 sent after scoring, 10 replies, 19 in a second round of follow-up.
Between 809 and 64, 745 leads went nowhere. That is the scoring doing its job, not a leak: the brief did not match what we do, the stated budget was too low, the poster had already closed it, or the same company had reposted with different wording. A person used to read all of those; now a person reads the 64.
What a 16% reply rate says, and what it does not
64 sent, 10 replies, 16%. That number says one thing: emails that were scored and drafted one at a time get answered. It says nothing about deals closed, and nothing about whether the rate holds in another industry. We ran no control group, so we do not claim it beats a person by any margin.
The other number that does not show is the rule that a failed lead is never marked done. A send error, a channel rejection, a dead address: the lead goes back into the queue for the next round or for a person. Without that rule leads vanish quietly, and you find out at reconciliation time.
The step that is still a person
The last look before sending. The drafts pick up details from the posting, and now and then get a detail wrong or overdo a courtesy. A person spends a few dozen seconds per email; those seconds stop a mistake going out under our name. We do not intend to automate that step.
Replies are a person too. The system marks that a reply arrived; what to say is decided by a person. The boundary sits there because the moment someone writes back is when the actual conversation starts.
For a business setting up lead automation
- Write the rule for not sending before the template for sendingThe leads the score stops are where the time goes; the template only shapes the small part that goes out.
- Keep a person on the last look before sendA few dozen seconds per email, and it stops mistakes going out under your name.
- Never let a failed lead be marked doneBack to the queue, retried, or handed to a person; otherwise it vanishes quietly.
- Measure reply rate against sent, not against fetchedFetched as the denominator turns a healthy number into a fraction of a percent.
- Record every stepTwo months on you should be able to answer why a given lead was never sent.
Tell us which step costs you the most
Contact us